Environmental Processes of the Ice Age: Land, Oceans, Glaciers (EPILOG) Alan C

Quaternary Science Reviews 20 (2001) 627}657

Environmental processes of the ice age: land, oceans, glaciers (EPILOG) Alan C. Mix! *, Edouard Bard", Ralph Schneider# !College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331-5503, USA "CEREGE, Aix-en-Provence, France #University of Bremen, Bremen, Germany

Abstract

Knowledge of the state of the earth at the Last Glacial Maximum (LGM, an interval around 21,000 years ago) is an important benchmark for understanding the sensitivity of global environmental systems to change. Much progress in understanding climates of the LGM has occurred in the &20 years since the end of the CLIMAP project of the 1970s (Climate Long-range Investigation, Mapping and Prediction). Here we review this progress, based on presentations and discussion at a "rst open science meeting of the EPILOG project (Environmental Processes of the Ice age: Land, Oceans, Glaciers) held in Delmenhorst, Germany, May 1999. We outline key controversies and document protocols for EPILOG contributions, so that a new synthesis of the LGM Earth can emerge as an open project of the world's community of scientists. ( 2001 Elsevier Science Ltd. All rights reserved.

1. Introduction looked or poorly represented processes that must be understood and incorporated into the next generation of A long-standing goal of paleoclimatologists is to models. At present, con"dence in model predictions of understand climate changes of the last ice age, the largest future climate change is limited, especially because many manifestation of natural climate change that remains of the models incorporate oceanic processes in a limited relatively well preserved in the geological record. This way. interest is more than just academic curiosity. Current If the geologic record is to provide a useful test of debate about the extent of future climate changes, climate models, we must "rst have con"dence that we whether natural or manmade, demands that we under- know what really happened during the last ice age. How stand the extent, rate, and frequency of natural climate cold was it? Was cooling relative to modern climate changes of the past. uniform over the globe, or do geographic patterns of Assessing the likelihood of future changes requires the change reveal the processes responsible for change? What use of computer models that link components to simulate mechanisms set the sensitivity of Earth's climate system the atmosphere, ocean, land surface, ice, and biota. In to large-scale changes? spite of growing sophistication of such e!orts, models inevitably simplify reality and must be tuned to simulate both the present system and its sensitivity to change. This 2. The CLIMAP reconstruction limitation of models underscores the importance of reconstructing and understanding the state of the Earth The challenge of reconstructing the surface of the during the Last Glacial Maximum (LGM). If models are Earth at the LGM was taken up nearly 30 years ago by able to simulate the full range of dynamic climate cha- the CLIMAP (Climate Long-range Investigation, Map- nges actually observed, then we gain con"dence in their ping, And Prediction) project, which built on a method ability to predict the future. If models fail in their simula- for quantitative estimation of past environments de- tions of the past, we learn from the failure about over- veloped by Imbrie and Kipp (1971). During the 1970s, the CLIMAP group pioneered a "rst global reconstruction of the ice-age climate. Along the way they revolutionized * Corresponding author. Tel.: #1-541-737-5212. the study of global environments in the geologic record E-mail address: [email protected] (A.C. Mix). as they developed strategies for quantitative stratigraphy

Fig. 1. Changes in the annual average sea surface temperatures (LGM-modern), with geographic distributions of LGM seasonal and perennial sea ice, and LGM land ice, based on seasonal estimates of CLIMAP (1981). A major conclusion of CLIMAP, surprising at the time, was that cooling associated with ice ages was regional, rather than global.

on a global scale, established a preliminary chronology Africa) cooled signi"cantly. Lesser cooling was inferred for global climate change, and advanced methods to near the equator in all oceans. The centers of the sub- estimate quantitatively seasonal temperatures based on tropical gyres and the western Paci"c warm pool yielded distributions of fossil species (e.g., Cline and Hays, 1976). the largest surprise; based on CLIMAP's species-based The CLIMAP project culminated in a series of maps methods temperatures in these areas stayed essentially that depicted the state of the Earth's surface at the LGM the same as (and in some cases even warmed relative to) (CLIMAP, 1981). Although these maps emphasized sea- modern conditions. Seasonal contrast increased near the sonal sea-surface temperatures, they also included the equator (implying stronger seasonal trade or monsoon distribution of ice both on land and in the ocean, as well winds), but decreased at high latitudes (due to perennial as vegetation and albedo (re#ectivity) of the land surface. ice cover on land and in the sea). Overall, CLIMAP A primary result of CLIMAP, surprising at the time, (1981) inferred that Earth cooled surprisingly little during was that the magnitude and even direction of temper- the ice age, except at high latitudes. This concept, often ature change was very di!erent between regions rather referred to as `polar ampli"cationa of climate change, than uniform globally (Fig. 1). The largest inferred chan- implies a feedback mechanism driven by increased al- ges were in the mid-to-high latitudes, especially in the bedo associated with expanded snow and ice cover. North Atlantic region bordered by large ice sheets. In the On land, forests shrunk, deserts expanded, and albedo CLIMAP reconstruction, sea ice advanced in the Green- was adjusted accordingly (Peterson et al., 1979). Debate land and Norwegian Seas and persisted through the within the glaciological community left uncertainty summer. Seasonal meltback of sea ice was also dimin- about the extent and thickness of LGM ice sheets. ished in the Southern Ocean. Upwelling systems asso- CLIMAP proposed two options, both based on equilib- ciated with eastern boundary currents (such as o! NW rium pro"les along presumed glacier #owlines (Hughes A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 629 et al., 1981). A `maximuma model assumed areas of ice As yet, no large-scale re-analysis of the LGM interval sheets as large as possible based on the greatest extent of exists based on recent estimates of ocean temperature, ice moraines or other evidence (in some cases undated) and sheets, and other environmental properties. Recognizing on an interpretation of ice #owlines that yielded a rela- the need for such a synthesis, 57 scientists met in May tively high Laurentide Ice Sheet dominated by a central 1999 at the Hanse Institute for Advanced Study in Del- dome. Signi"cantly expanded marine ice included an menhorst, Germany, to initiate the EPILOG program. Arctic ice shelf that helped to buttress thick ice on land. Research programs from nine nations (Australia, France, A `minimuma model was based on much more restricted Germany, Netherlands, Norway, Russia, Spain, UK, US) ice limits in many places, particularly around shallow were represented, and the e!ort was sponsored by marine basins, and on #owlines inferred from a di!erent IGBP/PAGES and IMAGES. This "rst public gathering interpretation of the glacial geology that in some cases of EPILOG set as its goals: (1) to assess advances made yielded thinner ice sheets. Hughes et al. (1981) acknow- in the 20 years since CLIMAP, "nding points of consen- ledged that real ice sheets, which respond slowly to cli- sus where possible, (2) to outline strategies that will help mate forcing, may not have achieved full equilibrium to resolve remaining con#icts, (3) to set standards for pro"les during the LGM, but given the data available at international contributions to a new community assess- the time there was no reliable basis for reconstructing ment of the LGM, and (4) to outline a plan of action transient behavior of the global ice sheet system. Indeed, leading to a new synthesis. The meeting emphasized Hughes et al. (1981) stated, `Our most important con- current understanding of sea-surface and continental clusion is that the distribution of late Wisconsin-Weich- temperatures at the LGM. selian ice sheets is not well knowna. This remains an Here we report on the views presented at this "rst important consideration, because climate feedback pro- EPILOG meeting. We focus on the rationale for a new cesses associated with ice depend on the regional details synthesis of LGM environments. We review substantial of ice extent and height (Ramstein and Joussaume, 1995; advances in chronology that help to re"ne research strat- Broccoli, 2000). egy and note key chronology issues that remain unresolved. We assess strengths and weaknesses of various geologic proxies for surface temperature and other in- 3. Two decades since CLIMAP formation needed for understanding ice-age climate, and note emerging regional syntheses based on new data. Our Over the past 20 years, several elements of the primary motivation is to highlight key controversies that CLIMAP (1981) reconstruction have been called into need resolution, to outline opportunities for research to question. Ice sheets based on regional sea-level data are "ll gaps in knowledge, and to document protocols for thinner and have about 35% less change relative to EPILOG contributions so that a new synthesis can modern ice sheets than those of the CLIMAP maximum emerge as an open project of the world's community of ice sheet and 10% less than the minimum ice sheet scientists interested in the evolution of past and future reconstruction (Fleming et al., 1998; Peltier, 1994, 1998a, climate. b; Lambeck et al., 2000), although recent data and isostatic modeling suggest a total volume change similar to the CLIMAP minimum ice sheet (Yokoyama et al., 2000). 4. Why the Last Glacial Maximum (LGM)? In the high-latitude oceans, diatom #oras appear to con- #ict with CLIMAP's (1981) inference of perennial sea ice The LGM is an important topic for paleoclimatic (Burckle et al., 1982; Crosta et al., 1998a). In the tropics, study today, as it was 30 years ago at the start of many climate models cannot reconcile CLIMAP's rela- CLIMAP, for many of the same reasons: tively small changes in tropical sea-surface temperature with evidence for advances of mountain glaciers, modi- E The LGM represents a global climate state dramati- "ed pollen assemblages (Rind and Peteet, 1985; Thom- cally di!erent from that of today, and thus provides pson et al., 1995) or rare gases in lowland groundwater a useful test of climate models' sensitivity to change. (Stute et al., 1995). New methods for reconstructing Although other extreme states of climate may also ocean temperature and other properties are being de- exist, the clearly documented in#uence of global ice veloped, but in some cases these con#ict with each other. sheets on climate (Clark et al., 1999) justi"es a thor- For example, Sr/Ca measurements in coral reefs suggest ough understanding of the climate state dominated by much greater cooling than that inferred by CLIMAP ice. (Guilderson et al., 1994), but an organic geochemical E The LGM is reasonably close to an equilibrium state thermometer based on unsaturation of long-chain al- of climate (to the extent that equilibrium is ever ob- Y kenones, U, suggest tropical temperature changes of tained in the recent geologic past); short-term variabil- about 23C that are to "rst approximation in agreement ity appears to be smaller during the glacial maximum with CLIMAP's results (e.g., Bard et al., 1997). than in older intervals such as marine oxygen isotope 630 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

stage 3, which is dominated by numerous short-term parison Project (PMIP). A summary of these model runs climate oscillations. (Pinot et al., 1999) supports earlier inferences that E Primary boundary conditions for the LGM climate, CLIMAP LGM SSTs may have underestimated ice-age such as continental geography, orbital con"gurations, cooling in the tropics, but also reveals that LGM cooling and atmospheric pCO are extremely well known. is clearly more pronounced over land than over the Others, such as sea level, ice area, and ice-sheet heights oceans for all models. CLIMAP (1981) estimated that are reasonably well known, and re"nements are average tropical SSTs cooled by 1.43C at the LGM. In imminent. the eight PMIP models that used these SSTs as a bound- E The LGM interval is within the range of both C and ary condition, atmospheric cooling at the sea surface was U-Th dating, and additional methods such as surface 0.8$0.13C in the tropics, but over land the average exposure dating with cosmogenic nuclides o!er the cooling was 2.4$0.43C. In the eight models that potential for uni"ed chronologies on land and in the simulated SSTs with simple `mixed-layera oceans, aver- sea. Major advances in this regard have been made age cooling over the sea surface was 2.1$1.03C, while since the CLIMAP study, including calibration of the over the land surface it was 3.4$1.23C. However, not all radiocarbon timescale to so-called calendar ages, and the models agreed about the overall magnitude of tem- the advent of accelerator measurements of C and perature changes. For example, LGM cooling over the other cosmogenic isotopes in extremely small samples. tropical continents varied between 1.8 and 5.43Cindi!er- E An extensive set of LGM geologic data and model ent models, implying that GCMs could be further im- runs based on both CLIMAP and post-CLIMAP proved. A key element that all of these models lack is an studies already exists. Re"nements to such work will interactive ocean that includes processes such as ocean be e$cient and cost e!ective. currents and upwelling of cool, subsurface waters. E Tremendous growth in computing technologies has Experiments using the current generation of coupled advanced the state-of-the-art in climate modeling, par- or partially coupled atmosphere}ocean models under- ticularly with the development of coupled atmo- score the importance of re-assessing the state of the LGM sphere-ocean models with a range of resolution and world, including the circulation and properties of the complexity. deep sea. Here we highlight three of these e!orts, along with the questions that arise from their results. The relatively few numerical climate models that exis- Ganopolski et al. (1998) considered a model with a dy- ted at the time of CLIMAP were limited to simulations of namical three-dimensional atmosphere of low spatial res- ` a perpetual seasons (Gates, 1976; Manabe and Hahn, olution (103 latitude by 513 longitude grid spacing, 16 1977; Kutzbach and Guetter, 1986). This modeling lim- vertical levels) coupled synchronously to a zonally aver- itation helped to motivate CLIMAP's experimental aged multi-basin ocean (Atlantic, Paci"c, and Indian ` strategy of estimating boundary conditions of Febru- Oceans). Ocean currents were parameterized based on a ` a ary and August sea-surface temperatures, in spite of latitudinal vorticity balance and meridional component questions about the statistical independence of such esti- of Ekman transport. The model calculated sea ice ther- mates. In contrast, today's models allow simulation of modynamically as the fraction of ice cover in each grid full seasonal cycles that include more realistic feedback cell and transported it using advection and di!usion. On processes such as moisture and vegetation feedbacks at land, vegetation was prescribed, and runo! was the land surface, more precise depictions of topography simulated by assigning excess moisture accumulating at of the land and the sea #oor, and elements of an interac- each land grid point to an appropriate point in one of the tive ocean. Some models can now be run through thou- two-dimensional ocean basins. The simplicity of this sands of years of simulation, allowing exploration of model allowed for long model runs (5000 years) ap- multiple stable states of the climate system and better proaching equilibrium conditions even in the deep sea. tests for the statistical signi"cance of relatively small Model output consisted of an average of the last 100 changes. With this growing sophistication of modeling, years of each simulation. an opportunity arises to use paleoclimate data in a new The LGM model run of Ganopolski et al. (1998) pro- way for model veri"cation, rather than as "xed boundary duced large atmospheric cooling at high latitudes, espe- conditions. cially in winter in the northern hemisphere (Fig. 2a). The tropical oceans cooled less, 2.33C on average (Fig. 2b), 4.1. Models and mechanisms while the average land surface in the tropics cooled by 4.63C. Near the equator the greatest atmospheric cooling, Recently, an array of 16 di!erent atmospheric general 4}53C, occurred over the Atlantic Ocean and the eastern- circulation models (some with mixed-layer ocean compo- most Paci"c Ocean (although the model ocean is two nents) was tested using CLIMAP LGM conditions (SST dimensional and thus does not really have east-west as either a boundary condition or as a veri"cation data structure). In the deep sea (Fig. 2c and d), the total rate of set) as part of the Paleoclimate Modeling Intercom- thermohaline overturn stayed about the same, but in the A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 631

LGM run the site of deepwater formation in the North Atlantic shifted southwards by about 203 of latitude (two grid boxes in the model). This southward shift of oceanic heat transport, accompanied by advance of sea ice in the North Atlantic, accounted for about half of the large climate response in the north (CO and land ice accounted for the rest). In the Paci"c Ocean, limited modern formation of North Paci"c Intermediate Water expanded to reach 2-km depth and this water mass crossed the equator to ventilate parts of the South Paci- "c. This expansion was driven primarily by cooling of the atmosphere at high northern latitudes, linked to the North Atlantic region via hemispheric winds. During the LGM the total rate of deepwater formation in this model was similar to the present rate in the North Atlantic and greater than at present in the North Paci"c; thus net northward heat transport increased. As a result of this northward heat transport, signi"cant cooling also occurred in the Southern Hemisphere, especially near 50}603S. Weaver et al. (1998) performed a similar experiment, but at a higher resolution (1.863 latitude by 3.753 longitude grid spacing), using a more complete three-dimensional ocean model. The 19-level primitive-equation ocean included components for thermodynamic sea ice and ice-albedo feedback, and was coupled to a simple energy and moisture balance model of the atmosphere (i.e., heat and moisture are transported via di!usion). In this model atmosphere, winds that drive ocean currents were approximated from the energy balance using a statistical scheme. The e!ect of atmospheric pCO was imposed by assuming that LGM reduction of pCO was equivalent to a 3.2 W m\ reduction in radiative forcing. Precipitation on land (which had no topography) re- turned to the ocean via 21 prescribed drainage basins. Length of runs was not speci"ed, but the authors stated that all runs reached equilibrium. In the Weaver et al. (1998) study, northern hemisphere cooling was strongest over North America and in the North Atlantic (Fig. 3a and b). This result is similar to that of Ganopolski et al. (1998) but was caused by a completely di!erent mechanism in which a nearly 50% reduction in the rate (but not the location) of North Atlantic Deep Water formation (Fig. 3c and d) caused a net

᭣&&&&&&&&&&&&&&&&&&&&&&&&&&&&&& Fig. 2. Results from coupled atmosphere}ocean models of Ganopolski et al. (1998); (a) Surface air cooling (LGM * modern control simulation). (b) Sea-surface temperature change (LGM * modern), zonal averages of the two-dimensional ocean basins. Hatched line is a slab- ocean model. Solid line is the coupled atmosphere}ocean model, and the dots are the CLIMAP reconstruction. Note that the coupled model produces substantial cooling ('83C) in the northern hemisphere. Both models reveal somewhat more tropical cooling than CLIMAP (1981). (c) Meridional overturning streamfunction, modern Atlantic basin, (d) as in c, but for LGM. Note the southward movement of the deepwater source area to &453N, and total Atlantic deep-water formation at LGM similar to today (&18 Sv). 632 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

reduction of northward heat transport (opposite the sense of change in the model of Ganopolski et al., 1998). As a result, oceanic heat built up in the South Atlantic and Southeast Paci"c sector of the Southern Ocean, mitigating regional cooling caused by lower greenhouse gases, and resulting in little or no net change in sea surface temperature in the Southern Hemisphere. Oceanic cooling in the model did occur near Australia, and in the North Paci"c (based on di!usion in the energy balance atmosphere from the Atlantic and North Ameri- can cold spot). Cooling in the tropical ocean was small (about 23C) and asymmetrical with latitude (greater cooling in the northern hemisphere) and longitude (greater cooling in the eastern reaches of each ocean basin). Cli- mate feedback mechanisms were analyzed by sequen- tially turning each one o!. Weaver et al. (1998) found that the most important mechanism for total cooling was radiative forcing associated with greenhouse gases, followed by ice-sheet albedo feedback. The e!ect of changing ocean circulation was primarily to change the spatial pattern of temperature changes. Bush and Philander (1998) used a fully coupled three- dimensional dynamical atmosphere}ocean model (2.253 latitude by 3.753 longitude, 14 atmosphere levels, and 15 ocean levels). The computational cost of this model limited the control run to 40 years, and the LGM run to 15 years, with the last six years used for output. This model run was too short to reach a full-equilibrium solution for the ocean interior, so results were only documented for the tropics and subtropics, where the dominant ocean circulation is shallow and the approach to equilibrium is rapid. The LGM Ocean is signi"cantly di!erent in the model of Bush and Philander (1998) that in those of Weaver et al. (1998) or Ganopolski et al. (1998). Cooling of the atmosphere at high latitudes compressed westerly wind belts equatorward in both hemispheres, and there- by intensi"ed both the westerly jets and the trade winds. As a result of greater subduction of cold water into the high-latitude thermocline and stronger equatorial upwelling, the tropical oceans cooled dramatically in the Bush and Philander (1998) model, for example by more than 63C in the western Paci"c. The similarities and di!erences between these three model runs are instructive both for the di$culties in- volved in understanding the LGM climate and for the opportunities that lie ahead. Although forced with

᭣&&&&&&&&&&&&&&&&&&&&&&&&&&&&&& Fig. 3. Results from coupled atmosphere-ocean model Weaver et al. (1998): (a) Surface air cooling (LGM * modern control simulation). (b) Sea-surface temperature change (LGM-modern); note limited cooling in the southern ocean Atlantic and eastern Paci"c sectors, a result of lower northward heat transport in the LGM. (c) Meridional overturning streamfunction, modern Atlantic basin, (d) as in (c), but for LGM. Note the stable location of the deepwater source area near 653N, with a reduction of Atlantic deep-water formation at LGM (&10 Sv) relative to today (&14 Sv). A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 633 similar boundary conditions (modi"ed radiation, re- the highest dO value, or in some cases the coldest duced greenhouse gases, continental ice sheets), the three inferred value with the general LGM interval. models give signi"cantly di!erent results and emphasize Signi"cant progress in geochronology and paleo- completely di!erent mechanisms. The mechanism noted climatology made during the 20 years following the by Ganopolski et al. (1998) is primarily oceanic (sea ice CLIMAP project allow us to revisit this issue. During the and a shift in deep-water formation), whereas that of "rst EPILOG meeting participants, assessed new evid- Weaver et al. (1998) is mostly atmospheric (CO and ence and reached an agreement on a practical time win- ice-sheet albedo). It is interesting to note that in each of dow for an LGM reconstruction. Three separate issues these models the primary mechanism that ampli"es arose: (1) establishing a clear de"nition of the LGM, (2) change occurs in the subsystem modeled in least detail. In assessing the precision and accuracy of LGM dates, and the Bush and Philander (1998) study, the strongest mech- (3) determining the timing and duration of LGM condi- anism is interaction of surface winds with the shallow tions on a global basis. ocean * a process that the other models do not include in detail. All three models highlight the importance of 5.1. Dexning the LGM based on proxy data understanding three-dimensional circulation of the ocean interior, because of the role deep waters play in control- We continue to de"ne the LGM as the time of the most ling large-scale heat transports from low to high latit- recent maximum in globally integrated ice volume. With- udes, or between hemispheres. Other primary processes in this interval, however, we seek a practical de"nition of are also possible in this highly linked system, such as the largest chronologic interval within which global cli- radiative feedback associated with water vapor (Webb mate was reasonably stable, while avoiding major cli- et al., 1997), and changes in the transports of fresh-water matic shifts or transients. Much work done within and vapor between ocean basins or from low to high latitudes since the CLIMAP project has con"rmed that regional (Zaucker et al., 1994; Rahmstorf, 1995; Hostetler and temperature records do not necessarily reach minimum Mix, 1999). values at the same time, so a de"nition based on the Con#icting results from models and di!erences in key coldest observed temperature or some other local ex- feedback mechanisms that set the sensitivity of each treme conditions would be inappropriate. Similarly, it is model to change beg the question, `what really happened unlikely that all ice sheets and glaciers on the earth during the LGM?a Before further progress can be made reached a synchronous maximum. The challenge is thus in modeling and understanding the mechanisms of ice- to characterize unambiguously the maximum of conti- age climate change, a new comprehensive reconstruction nental ice volume in paleoclimatic archives that integrate of LGM climate is needed. This reconstruction can then global e!ects, to establish the timing of such e!ects, and be used to test whether mechanisms suggested by the to "nd ways of linking various local or regional climate models are reasonable or not. records into this global framework. Two independent methods are commonly used to assess globally integrated ice volume. First, the oxygen d 5. When was the Last Glacial Maximum (LGM)? isotopic composition of the mean ocean water, O , changes in response to storage of O-depleted ice on The LGM can be de"ned as the most recent interval land (Emiliani, 1955; Shackleton, 1967). dO is 1000 when global ice sheets reached their maximum integrated ( O/ O! %/ O/ O%&%%#% }1) and reference is Stan- volume during the last glaciation. When did this occur? dard Mean Ocean Water (SMOW) for water or atmo- Based on stratigraphy available in the early 1970s, spheric oxygen samples, or Pee Dee Belemnite (PDB) for CLIMAP (1976) selected the celebrated benchmark of calcite samples. Second, sea level changes relative to the 18,000 C-yr BP and assumed climate stability for the modern baseline as water is removed from the ocean to interval between 14,000 and 24,000 C-yr BP. Not all build ice sheets. This portion of sea level change is called records considered by CLIMAP included radiocarbon equivalent sea level (ESL), which is inferred from various dates. In many cases, CLIMAP cores were dated by dO local records of relative sea level (RSL) by modeling the stratigraphy, in which the LGM was de"ned as the mid- gravitational e!ects of loading by ice on land or water in point of oxygen isotope stage 2 (Shackleton, 1977). In the ocean basins. d d some cases O data were not available and the stage Changes of O have been inferred from di!erent 2 boundaries were inferred based on some other regional archives such as planktonic and benthic foraminifera lithostratigraphy or biostratigraphy, such as %CaCO (Mix and Ruddiman, 1985; Shackleton, 1987; Labeyrie (McIntyre et al., 1976) or the percentage of indicator et al., 1987; Mix, 1987), deep-sea sediment pore waters species such as the radiolarian Cycladophora davisiana (Schrag and DePaolo, 1993; Schrag et al., 1996) and (Hays et al., 1976). Within that broad interval, a represen- molecular oxygen from atmospheric bubbles trapped in d tative estimate was chosen based on an average of values polar ice ( O! , Sowers and Bender, 1995). To a "rst within the LGM interval, or a single value at the depth of approximation the three techniques provide the same 634 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

d basic information, i.e., O was signi"cantly higher former ice sheets. Using sediment data and box models, during the LGM than at present. However, ever since the several authors suggested that the dO record of the pioneering work of Emiliani (1955) and Shackleton deep Paci"c could have lagged that of the Atlantic signi"- (1967), uncertainties about the exact magnitude of cha- cantly, depending on the deep-sea circulation rate that d d nges in O have been debated. This is because O prevailed in the past, although such transient e!ects are d from each archive combines changes in O with most prevalent in the Atlantic ocean where most of other sources of isotopic variability. glacial meltwater entered the ocean (Broecker et al., 1988; Planktonic foraminifera dO curves are a!ected by Bard et al., 1991; Duplessy et al., 1991). A "nding that the local changes in ambient temperature, #uctuations of the dO shift following LGM conditions in the deep Paci"c regional balance of evaporation and precipitation, and appears to precede rapid changes in sea level (Mix et al., transport of fresh water via atmospheric vapor between 1999a) suggests that deep-sea warming may have preced- oceans or from low to high latitudes (Broecker, 1986; ed major deglaciation (meltwater pulse 1A as dated by Mix, 1987). Near coastal sites, river runo! with dO Bard et al., 1990a). An alternate scenario, based on a new values lower than local seawater adds additional com- "nding of rapid sea-level rise near 19,000 cal-yr BP, may plexity (Pastouret et al., 1978). Local or regional suggest that isotopic change in the deep sea lags behind anomalies that decouple planktonic dO values from the termination of the LGM interval (Clark and Mix, global ice volume e!ects also include glacial meltwater 2000; Yokoyama et al., 2000). that enters the ocean (Kennett and Shackleton, 1975; Another isotopic method of assessing isotope budget Cortijo et al., 1997). Given all of these regional complica- changes, based on the oxygen isotopic composition of tions, as much as one-third or even half of the total molecular oxygen in air bubbles preserved in ice cores d d variability in planktonic foraminiferal O curves may (here, O! ) also does not record the pure oceanic d d re#ect local environmental changes, which do not neces- component of O variations. Changes in O! sarily mimic those of ice volume. also record the balance of marine and terrestrial produc- Benthic foraminifera inhabit the deep sea, which is tion and consumption of oxygen by photosynthesis and generally more stable and uniform in its properties than respiration (the `Dole E!ecta; Dole and Jenks, 1944). are surface waters. Nevertheless, it is now recognized that They may also re#ect divergence of surface ocean dO most benthic foraminiferal dO curves, with total ranges from that of the mean ocean, due to changes in the rate of between 1.5 and 1.9&, are also a!ected by signi"cant isotopically depleted water vapor from low to high latit- deep-sea temperature changes. Estimates of abyssal tem- udes where it is preferentially transported into the deep perature changes vary from 23C (Shackleton, 1987), to sea (Broecker, 1986; Mix, 1992). The observed maximum d d 4.53C (Dwyer et al., 1993), which is equivalent to a O of O! during the last ice age occurred very late (i.e., range in calcite samples of 0.5}1.1&, respectively. Con- 15,000 cal-yr BP) in the well-dated GISP2 ice core sidering likely changes in temperature and other local (Sowers and Bender, 1995) relative to other estimates of e!ects, foraminiferal dO data suggest a glacial-inter- the glacial maximum (near 21,000 cal-yr BP, see below) glacial change due to ice volume of 1.1}1.4&. This is and the extent of change from 15,000 to 8,000 cal-yr BP is roughly equivalent to excess LGM ice volume of quite large (1.5%; Sowers and Bender, 1995). In addition, ; d 48}59 10 km . Here we assume modern ice volume of the O! record generated for the four last glacial 32;10 km, a mean dO of modern glacier ice (includ- cycles in the Vostok ice core (Petit et al., 1999) is clearly ing Antarctica) of !47& relative to SMOW, and mean dominated by a &22-kyr cycle, which contrasts with the dO of LGM ice (including Antarctica) of !44& rela- classical 100-kyr cycle thought to dominate Late-Quater- tive to SMOW (!37& for ice outside of Antarctica). nary ice-age oscillations (Imbrie et al., 1993). Thus, it The total ice volume implied by these assumptions is appears likely that regional climatic or biotic e!ects shift d smaller than the maximum ice sheet reconstruction of the timing and pattern of changes in O! recorded in Hughes et al. (1981), but near that of the minimum model ice cores relative to that of global ice volume. This limits ; d (52 10 km more ice than at present). For comparison, the use of O! in determining the timing and ampli- recent excess LGM ice volume estimates based on tude of the LGM. modeling of sealevel data range from 40;10 km Schrag and DePaolo (1993) and Schrag et al. (1996) (Peltier, 1994) to 52;10 km (Yokoyama et al., 2000). used direct measurements of dO in deep-sea sediment d Changes in the isotopic composition of glacier ice pore waters to estimate the O during the LGM. through time suggest the possibility that changes in This technique, in which the record is deconvolved from d O may lag changes in ice volume by as much as data smoothed by di!usion, suggests a much lower range a few thousand years during major transitions (Mix and of dO changes (0.8}1.0&, Schrag et al., 1996) than that Ruddiman, 1984). In addition, the deep ocean was prob- based on temperature-corrected foraminifera or corals ably not isotopically homogenous. The in#uence of O- (1.1}1.4&, Labeyrie et al., 1987; Mix and Pisias, 1988; depleted meltwater during the deglaciation probably "rst Fairbanks, 1989), or that implied by recent sealevel re- reached the Atlantic Basin, which was closest to the constructions (Clark and Mix, 2000; Yokoyama et al., A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 635

2000). The cause of this mis"t is not presently under- 5.2. Accuracy for dating the LGM interval stood. Future e!orts to analyze chlorinity and hydrogen isotope composition of the same pore water samples may Radiometric dates for LGM events are most com- help to constrain the interpretation of porewater data monly based on radiocarbon. These radiometric dates (D. Schrag, pers. comm., 2000). Although useful for can be adjusted to provide an estimate of calendar ages d assessing the amplitude of change in O , the by correcting for changes in C production rates and porewater method does not provide precise information some reservoir e!ects. Calendar calibrations now extend on the timing of the LGM, and is not useful as a tool to through the LGM to 20,400 C-yr BP (equivalent to locate LGM sediment samples, because the signal is 24,000 cal-yr BP in CALIB-4), based on a combination of heavily smoothed and transported from its original C-dated annually layered materials such as tree rings, stratigraphic position by di!usion of pore waters within as well as marine carbonates from the near-surface ocean the sediment column. dated with both C and U-Th (Stuiver et al., 1998; Bard Perhaps the most direct way to detect the glacial et al., 1998). A software program commonly used maximum is by documenting the low-stand of sealevel for calendar corrections of radiocarbon dates, CALIB, caused by the lockup of water on land-based ice sheets. It updated recently to version 4.2 (http://depts.washing- is possible to document deglacial sea levels by drilling ton.edu/qil/calib), is recommended as the standard cor- tropical reefs and coastal sediments. To a "rst approxi- rection scheme for EPILOG contributions. mation all relative sea-level curves track the global sea- CALIB uses a simple atmosphere}ocean box model to level changes that occurred since the LGM. However, propagate inferred changes in atmospheric production of second-order #uctuations superimposed on the common radiocarbon into the surface and deep-sea reservoirs, and eustatic change are linked to regional and local e!ects is well constrained at accepted calibration points, but it (see atlas compiled by Pirazzoli, 1991). This complexity does not explicitly assess changes in ocean circulation justi"es the use of geophysical models which take into that may change the regional distribution of C within account all data on sea level and on the areal retreat of the oceans. Thus, CALIB corrections are most accurate continental ice sheets, including comparison of sea-level when applied to pristine samples from land (such as records far from the ice sheets (Edwards et al., 1993; Bard wood) or carbonate shells or corals that grew in the et al., 1996) with those closer to the ice sheets (Fairbanks, near-surface waters of the low-latitude ocean that are 1989; Bard et al., 1990b). Results based on this approach near equilibrium with the atmosphere. Calendar correc- range from a total equivalent sea-level change of about tions in the deep sea or polar oceans are less certain, due 105 m (Peltier, 1994), to 120}130 m (Fleming et al., 1998; to the large regional (and possibly temporal) variations in Peltier, 1998a and b). However, new data from sediments reservoir ages (e.g., Shackleton et al., 1988; Bard et al., on the Australian margin suggest a total equivalent 1994; Broecker et al., 1998; Sikes et al., 2000). Correc- sea level lowering of about 135 m (Yokoyama et al., tions for the LGM interval are less certain than in the 2000), which implies excess LGM ice volume of Holocene, due to the limited number of well-constrained 52$2;10 m. At present, no high-resolution sea-level calibration points. It is likely that additional details, such curves based on dated corals extend through the full as radiocarbon `plateausa will emerge (e.g., Kitigawa and range of the LGM. Sea-level reconstructions consider van der Plicht, 1998), but inclusion of such e!ects is only land ice, as #oating ice shelves would displace premature at present. water and thus would yield little or no change in sea Direct dating of corals using both C and U-Th level. methods provides the best available estimate of the true For purposes of de"ning the LGM interval, the EPI- age of the LGM based on dating the interval of lowest LOG participants agreed that local glacial maxima or sealevel. The scarcity of LGM samples recovered from temperature minima do not provide a reliable strati- coral reefs and shallow-water sediments, however, is graphic link to the globally integrated glacial maximum. a major obstacle to this approach. Sites lacking major The best estimate of such an integrated maximum is complications of tectonic uplift are now submerged and based on the equivalent sea level low stand (i.e., after thus can only be sampled using ocean drilling * an removing local isostatic e!ects), followed by benthic expensive and logistically complicated proposition. foraminiferal dO (recognizing the presence of some Existing coral data (Fig. 4) reveal rapid sea-level rise temperature or watermass e!ects even in the deep sea). near 12,000 C-yr BP (14,000 cal-yr BP). This event, Because various components of the geologic record will commonly referred to as Meltwater Pulse 1A (MWP-1A, not contain either of these measures, however, the most Fairbanks, 1989) postdates the beginning of the B+lling practical de"nition of the LGM horizon for purposes of warm event in Europe (near 12,600 C-yr BP, or mapping is one based on chronology. Links into this 14,800 cal-yr BP). Recent data from organic debris o! the de"nition will inevitably use a variety of stratigraphic Sunda Shelf in Southeast Asia are consistent with rapid and chronologic approaches. This raises the question of sea-level rise, and radiocarbon dates here suggest timing accuracy of dates for the LGM. between 14,100 and 14,700 cal-yr BP (Hanebuth et al., 636 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

2000). U}Th dates at Barbados (which require no reservoir or calendar corrections as do the C dates) constrain rapid sea-level rise of MWP-1A to younger than 14,300 cal-yr BP (Bard et al., 1990a). The lowest stand of sea-level occurred prior to 16,000 C-yr BP (19,000 cal-yr BP) (Fairbanks, 1989; Bard et al., 1990a, b) and probably around 18,000}19,000 C- yr BP, or 21,200}22,400 cal-yr BP (Fleming et al., 1998). Yokoyama et al. (2000) provide new evidence from the tectonically stable Australian margin for a rapid sealevel rise near 19,000 cal-yr BP, which is preceded by a relatively stable 3000}4000 yr interval of an equivalent sealevel low stand &135 m below present. If con"rmed, this early sea-level rise provides a logical younger limit to a well-de"ned LGM, centered at 21,000 cal-yr BP and extending from 19,000 to at least 23,000 cal-yr BP. This de"nition of the LGM interval is remarkably close in time to the original de"nition of CLIMAP (1976, 1981) of 18,000 C-yr BP. One recommendation of the EPILOG group is that this LGM sea-level low stand must be understood better, through e!orts to drill, sample, and date submerged coral reefs and shallow-water sediments.

5.3. Transient events and regional timing of the LGM conditions

Was the LGM a very brief event or an extended period of stable climate? This question is crucial as it a!ects the possibility of reconstructing and simulating the LGM as an equilibrium climate state. There is also a practical reason for evaluating the duration and regional timing of LGM conditions; geologic records are generally obtained as discrete data points rather than continuous curves, Fig. 4. Equivalent sea level (ESL) estimates based on coral data corrected for global isostatic e!ects (K. Lambeck, personal communication, and it is rarely possible to assign exact dates to each data 2000) compared to deep Paci"c benthic foraminifera dO (Mix et al., point. Thus, a useful de"nition of an LGM interval is 1999a). Black triangles are based on data from Barbados (Fairbanks, needed to allow for reasonable error in dates, and for 1989; Bard et al., 1990b), red squares are based on data from New appropriate averaging of paleoclimatic proxies with dif- Guinea (Edwards et al., 1993), and blue circles are based on data from ferent time resolutions. In addition, transitions into and Tahiti (Bard et al., 1996). Open circles are isostatically corrected sediment facies data from northern Australia (Yokoyama et al., 2000). The out of the LGM climate state may have occurred at solid black line is a 1000-year block average of the ESL data, calculated di!erent times in various places on the globe. A "nal every 500 years. Note rapid ESL rise near 19,000 cal-yr BP, which chronologic de"nition of the LGM should be based on terminates the LGM interval. The gray line with small dots is a deep calendar ages, and should be equally applicable (i.e., d d Paci"c O record based on benthic foraminifera. Note rapid O avoiding regional transient climatic shifts) in as many change near 16,000 cal-yr BP, suggesting abyssal warming that postdates initial sealevel rise. The two panels span a range of possible scales areas as possible. for the relationship of ESL dO in the deep Paci"c. In panel (a) the Although sea-level data o!er the best available inte- sealevel and isotope records are scaled such that 1 m change of ESL is gration of the global ice volume, they are less useful in associated with 0.011& change of dO (Fairbanks and Matthews, assessing "ne-scale variability within the LGM interval, d 1978). This scaling implies a total glacial-interglacial O change due and in many cases cannot assess regional di!erences in to ice volume of 1.5&, limited deep-Paci"c cooling of about 13C early in the deglaciation, and warming to essentially modern conditions near timing of major events. Since the CLIMAP (1976, 1981) 16,000 cal-yr BP. In panel (b) the sealevel and isotope records are scaled studies, many high-resolution records have become avail- such that 1 m change of ESL is associated with 0.0075& change of able from ice cores and marine sediments that o!er dO. This scaling implies a total glacial-interglacial dO change of guidance about choosing a practical LGM interval. & 1.0 (Schrag et al., 1996), substantial deep-Paci"c cooling of about 33C These show that the climate of the last ice age was far (relatively to modern bottom water temperature at the side of 1.73C), and continual warming throughout the deglacial interval in a series of more variable than previously thought. In particular, the steps near 16,000, 13,000 and 10,000 cal-yr BP. LGM does not always correspond to the coldest temperatures. For example, in the North Atlantic Ocean and A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 637 surrounding continents, temperature minima typically dO values are remarkably stable in the interval around occurred during Heinrich Events (HE) (e.g., Bard et al., the LGM (Fig. 5a). This is probably not an artifact of 2000). These events are de"ned based on dating of sand smoothing caused by bioturbation, as sedimentation layers (so-called Heinrich Layers) in North Atlantic sedi- rates are high, about 15 cm ka\ in this interval. How- ments, which are inferred to be caused by ice-rafting ever, the highest dO values occur near 18,000 cal-yr BP (Heinrich, 1988; Bond et al., 1992). The LGM, as de"ned and 26,000 cal-yr BP, both prior to and following the by sea-level records, probably took place in the interval LGM interval de"ned by the lowest stand of sealevel bracketed between the end of HE-2 (19,500 C-yr BP) near 21,000 cal-yr BP (Yokoyama et al., 2000). and the start of HE-1 (15,000 C-yr BP), two prominent The rapid change in benthic foraminiferal dO near events which have been dated precisely with accelerator 16,000 cal-yr BP appears to precede the rapid rise in mass spectrometry radiocarbon measurements on sea-level of MWP-1A near 14,000 cal-yr BP (Fig. 4) but foraminifera (Elliot et al., 1998; Thouveny et al., 2000). postdates the early sea-level rise near 19,000 cal-yr BP These dates correspond to calendar ages of about 23,000 (Yokoyama et al., 2000). If all these records are correct and 18,000 cal-yr BP, respectively. Global climate was there is some decoupling of benthic foraminiferal dO clearly not at steady state during these events of massive from sea level. The timing di!erences are larger than surges and melting of icebergs, which a!ected many com- transients lags induced by ocean mixing (Duplessy et al., ponents of the ocean}atmosphere}cryosphere}biosphere 1991). A solution may involve the timing of deep-sea system. The consensus of the EPILOG group is that temperature change (Mix et al., 1999a), and the dynamics Heinrich Events should be excluded as much as possible of ice shelves (which a!ect the isotope budget but not from an LGM reconstruction. sea-level) and land ice (which a!ect both the isotope Other important regional climatic events may also budget and sealevel), Clark and Mix (2000). Thus, al- need to be excluded from the LGM time window. For though these dO data do not support rapid transient example, the so-called Dansgaard/Oeschger (D/O) events of ice volume or deep-sea temperature during the Events, de"ned in Greenland ice cores, are brief episodes glacial maximum, the apparent decoupling of sea-level of warming in the North Atlantic Ocean and surround- and isotope records suggests that not all events are ing area. In particular, D/O-2, a prominent warm event, clearly recorded by dO. For purposes of de"ning the should probably be excluded from the LGM time win- LGM interval, however, a conclusion is that the highest dow. It remains uncertain to what extent D/O Events are value of benthic foraminiferal dO is not precisely alig- expressed throughout the globe, although it has been ned with the lowest sealevel stand, and may even be o!set suggested that equivalent warm events exist in the North in time by as much as a few thousand years. These Paci"c (Hendy and Kennett, 1999). di!erences, if con"rmed, should be considered when em- It is also important to consider the major regional ploying dO stratigraphy in the assembly of LGM data. climatic events that occurred in the Southern Hemi- The most precisely dated deglacial record of water sphere. Events of the high southern latitudes, including dO in ice (which mainly re#ects regional temperature of the end of extreme cold conditions associated with the precipitation) comes from the Byrd core (Johnsen et al., LGM, are thought to have led those of the northern 1972). Three independent techniques have been used to hemisphere (Imbrie et al., 1989; Charles et al., 1996; Clark date this record. First, the electrical conductivity method et al., 1996). Southern Hemisphere polar events are so far (ECM) provides stratigraphic information on annual best dated in ice cores from coastal and central Antarc- bands for layer counting (Hammer et al., 1994). Second, d tica (i.e., at the sites of Byrd, Vostok and Taylor Dome), the record of O! #uctuations in the Byrd core has which have been correlated to those of Greenland. been matched to that of the GISP2 ice core in Greenland Here (Fig. 5) we evaluate the potential for errors or (Sowers and Bender, 1995; Bender et al., 1999), which was biases in an LGM reconstruction associated with re- previously dated by layer counting (Alley et al., 1993). gional climatic changes or transient events, using well- Third, methane concentration data from the Byrd core dated high-resolution records of benthic foraminiferal have been correlated to similar #uctuations in the GRIP dO (North Paci"c core W9809A-13PC; Lund and Mix, ice core in Greenland (Blunier et al., 1998). d 1998, Mix et al., 1999a) and O*#% from the GISP2 and Correlation of the Byrd ice core to Greenland in the GRIP ice cores in Greenland (Grootes et al., 1993; John- LGM interval (and by inference the relative timing of sen et al., 1997) and from the Byrd ice core in Antarctica extreme cold conditions in the high latitudes of the (Johnsen et al., 1972). southern and northern hemispheres) remains unclear. d d The chronology for the deep-sea record of O from The variations of O! and CH in this interval are core W8709A-13PC is based on 41 AMS C ages (36 small and not uniquely diagnostic, which leads to poten- within the window 10}30 kyr cal BP) measured separate- tially signi"cant chronological errors in the LGM time ly on planktic and benthic foraminifera (Mix et al., range. Indeed, Sowers and Bender (1995) acknowledge ` d 1999a). Radiocarbon ages were corrected to calendar that between 16,000 and 22,000 cal-yr BP, O! var- ages using CALIB-4. In this record, benthic foraminiferal ied by less than $0.1 permil. Thus during this period, we 638 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

Fig. 5. Oxygen isotope time series for the time period 10}30 cal-kyr BP. Here, Hol is Holocene, YD is Younger-Dryas, A-B is Aller+d-B+lling, ACR is the Antarctic Cold Reversal, and DO2 is Dansgaard-Oeschger event 2. An1 refers to the last dO maximum before the LGM in the Byrd core from Antarctica. (a) dO in benthic foraminifera from the Paci"c core W8709A-13PC. The time scale is based on 41 AMS C ages (36 within the window 10}30 kyr cal BP) measured on planktic and benthic foraminifera (Mix et al., 1999a), corrected to calendar ages using CALIB-4.0. (b) dO in ice from the GISP2 core in Greenland (Stuiver and Grootes, 2000). The time scale is the GISP2 chronology available on the 1997 GRIP-GISP CD ROM ("le gispd18o.dat). The main reference for the age scale is Meese et al. (1994) but the CD-ROM "le includes the revisions by D.A. Meese as of September 1994, later published by Alley et al. (1997) and Meese et al. (1997). (c) dO in ice from the GRIP core in Greenland. The lower curve with left y-axis is plotted versus the &conventional' GRIP time-scale based on a #ow-model (Dansgaard et al., 1993, Johnsen et al., 1997); i.e., "les gripd18o.dat and gripage.dat of the GRIP-GISP CD-ROM). The upper curve with right y-axis shows the same dO data plotted versus the stratigraphic chronology by Hammer et al. (1997); i.e., "le gripstrt.dat of the 1997 GRIP-GISP CD-ROM. (d) dO in ice from the Byrd core in Antarctica (Johnsen et al., 1972). The lower curve with left y-axis is plotted with the time-scale by Sowers and Bender (1995) slightly modi"ed by Bender et al. (1999). The upper curve with right y-axis is plotted versus the time-scale of Blunier et al. (1998). Both Byrd chronologies are indirect and based on synchronizing geochemical d proxies #uctuations measured in air bubbles ( O! and CH) with the corresponding variations dated in Greenland ice. Consequently, the Byrd chronology by Bender et al. (1999) (lower curve in (d)) is compatible with the GISP2 chronology (b). Similarly, the Byrd chronology by Blunier et al. (1998) (upper curve in (d)) is compatible with the &conventional' GRIP chronology (lower curve in (c)). A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 639 are unable to provide well-constrained time scales for C and Taylor Dome suggests that deglacial warming at d a other ice cores using O! . Similarly, Blunier et al. Taylor Dome is older than previously thought and syn- (1998) state `the time period 17,000}25,000 cal-yr BP was chronous with other sites in Antarctica (Mulvaney et al., excluded from the synchronization because only low 2000). CH variations appear during that period, making the The Vostok ice core has a lower accumulation rate and correlation uncertaina. Recognizing the uncertainties im- thus a larger gas age ! ice age di!erence than the cores plied by these di$culties in correlating the Byrd ice core from Antarctica coastal sites. These properties lead to to Greenland, the coldest (lowest ice dO, Fig. 5d) inter- larger dating uncertainties associated with correlation to val in west Antarctica appears to occur between about the Greenland records. Indeed, discrepancies on the or- 19,500 and 22,000 cal-yr BP (Bender et al., 1999) or be- der of 2000}3000 years appear during the last deglaci- tween 20,000 and 22,500 cal-yr BP (Blunier et al., 1998). ation when comparing the Vostok chronology based on Ice dO values rise unambiguously above the range of correlation to the GISP2 chronology (Sowers and typical glacial values by 17,000 cal-yr BP. Bender, 1995; Steig et al., 1998) with those based on the Within the Greenland ice cores, the timing of LGM `conventionala GRIP chronology (Blunier et al., 1998). events based on the `conventionala GRIP and GISP2 For example, the level in the Vostok ice core dated at chronologies di!er by about 2000 yr (compare Fig. 5b 16,500 cal-yr BP by Blunier et al. (1997) is inferred to be and lower curve in 5c). It is thus remarkable that the two 19,700 cal-yr BP by Steig et al. (1998). These discrepan- Byrd chronologies developed by synchronization to cies illustrate the overall uncertainties in the dating pro- either GRIP (via CH, Blunier et al., 1998) or GISP2 (via cess (see also the discussion by Fischer et al., 1999). d O! , Bender et al., 1999) are quite close (Fig. 5d) even Nevertheless, it is still possible to conclude that the though the GRIP and GISP2 records in Greenland dis- deglacial warming (based on rise of dD) in Central Ant- agree by two millennia (Fig. 5b and c). This must be an arctica started somewhere between 18,500 and 21,500 d artifact. Correlating methane and O! in the Byrd cal-yr BP. The latter date would imply that transient core to the same two data types in GISP2 would shift the warming in Central Antarctica started during the LGM record near the LGM by nearly 3000 years. The problem interval de"ned by sea level. may lie in details of the records within the LGM interval. Given di!erences in the various ice cores, and the Consider the last dO maximum of the LGM interval at deep-sea records, it remains uncertain exactly when de- Byrd (An1), illustrated in Fig. 5d. Bender et al. (1999) date glacial warming began on a global basis. Clear di!er- this event at 22,300 cal-yr BP, while Blunier et al. (1998) ences in timing and character of the records exist between place it 900 years earlier, at 23,200 cal-yr BP. For com- Greenland and Antarctica, as well as between the various parison, in the GISP2 ice core DO2 occurs almost 1,000 records within Antarctica. Signi"cant dating uncertain- years earlier than An1 (Bender et al., 1999), while in the ties exist in the ice cores for the interval around the GRIP ice core DO2 occurs almost 2 kyr later than An1 glacial maximum. Nevertheless, we can summarize broad (Blunier et al., 1998). Further synchronization of the Byrd regional di!erences based on available data. In the North stratigraphy to GISP2 based on CH peaks now suggests Atlantic region, major warming was roughly coeval with an age for the ice dO minimum between ca. 21,000 and the start of the B+lling-Aller+d interstade, near 23,000 cal-yr BP (Blunier and Brook, personal commun- 14,600 cal-yr BP, and extreme cooling events appear to ication, 2000), slightly older than the ages of Bender et al. be associated with Heinrich Events 1 and 2 (17,000 and (1999). Clearly some puzzles remain to be solved regard- 25,000 cal-yr BP, respectively). In Antarctica, signi"cant ing ice-core chronologies near the LGM. deglacial warming began by 17,400 cal-yr BP, following The isotopic record from the Taylor Dome (Steig et al., initial warming that dates sometime between 19,500 cal- 1998) was dated by a combination of dating methods, yr BP (Byrd core) and 20,000$1500 cal-yr BP (Vostok including Be #uxes and synchronization of CH and core). Based on discussion at the EPILOG meeting, it d d O! records to those of Greenland. The D record was agreed that an LGM time slice should minimize the (i.e., the deuterium isotopic composition of ice) here is e!ects of this early warming in Antarctica, and avoid characterized by a rather late and abrupt deglacial warm- major regional transient events, if possible. ing near 14,800 cal-yr BP. This event appears to be synchronous with the start of the B+lling warm event as 5.4. Choosing an LGM chronozone dated in Greenland ice. At Taylor Dome, the interval between 20,000 and 15,000 cal-yr BP exhibits cold tem- Considering the sea-level constraints and the detailed peratures with limited variability (note, however, a warm records of regional climate change available from the ice event centered on 16,000 cal-yr BP). Unfortunately the cores, the EPILOG group reached a consensus that Taylor Dome record does not extend beyond 20,000 cal- a preferred LGM chronozone (referred to here as LGM yr BP. Thus it is not possible to assess the climate varia- Chronozone Level 1) can be de"ned as the interval be- bility at this site over the full range of the LGM. Further- tween 19,000 and 23,000 cal-yr BP (i.e., 16,100}19,500 more, a recent comparison of calcium records of Dome C-yr BP). This 4000-yr time window, centered on 640 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

21,000 cal-yr BP, encompasses the center of the LGM maximize the probability that an LGM reconstruction event de"ned previously by CLIMAP (1976, 1981), and is will approximate an equilibrium climate state. Our tests long enough to allow the inclusion of much existing examined the three data time series of O in Fig. 5 (one paleoclimatic data in a new synthesis. It is coeval with the oceanic, W8709A-13PC from the North Paci"c and two lowest stand of sea level (Yokoyama et al., 2000), avoids from ice cores, GISP2 from Greenland and the Byrd core all known Heinrich Events in the North Atlantic region, from Antarctica). These examples do not reveal all pos- and excludes most of Dansgaard-Oeschger climate event sible issues with data treatments, but we consider them D/O-2, as dated in the GISP2 ice core and in the GRIP representative of the typical high-resolution paleocli- core (with the chronology of Hammer et al., 1997). The matic time series that might be analyzed in di!erent parts time window avoids the major deglacial warming in of the world (whether based on isotope measurements or Antarctica, although it may include the beginning of the other data types). Our goal was to "nd data treatments earliest warm events in the south (Blunier et al., 1998; that return stable estimates of the `typicala value and Bender et al., 1999). This issue will require strengthening variability within the LGM chronozone, i.e., representa- of ice core dating to resolve the discrepancies among the tive values that are insensitive to realistic errors in chro- published studies. nology or di!erences in resolution likely to be found in The EPILOG group recognizes that realistic uncer- geologic records. tainties in dating must be accommodated in any time- As estimates of the typical LGM value within the slice reconstruction, and suggests a second option for the chronozone we considered (1) the data point closest to LGM chronozone (Level 2) between 18,000 and 21,000 cal-yr BP, (2) the arithmetic mean of all points 24,000 cal-yr BP (15,000}20,400 C-yr BP; Bard, 1999). within the chronozone, (3) the median, (4) the minimum, This is slightly broader than the preferred chronozone, (5) the maximum, (6) the midrange (halfway between the and may partly include transient events such as DO2 in minimum and maximum values), and (7) the trimmed the North Atlantic, and the start of deglacial warming in mean (Tukey, 1970). For the trimmed mean, the data the Antarctic. values in a sample are ordered, the highest 25% and Finally, recognizing that some records will have lim- lowest 25% of the values are removed, and the average of ited chronological control but may still provide useful the middle 50% of values was taken. This method is information we suggest a third, less restrictive, de"nition useful for removing outliers and reducing skewness in of LGM Chronozone Level 3. Our purpose in providing non-Gaussian or noisy data sets. As measures of variabil- these di!erent levels of certainty is to strike a balance ity within the LGM chronozone, we considered the range between reconstruction of a precise chronozone repre- and the standard deviation (without culling outliers). sentative of the global glacial maximum, and inclusion of We tested stability of each of these measures in each as much data and geographic coverage as possible. core to the time-slice de"nition (Chronozone Levels 1 and 2), to dating errors of $1000 years, and to data E LGM Chronozone Level 1 19,000}23,000 cal-yr BP. density. Dating errors were simulated by moving each Chronologic control based either on annually counted time slice younger or older by 1000 years. The e!ect of layers extending through the LGM chronozone, or sample density was assessed by removing samples from two radiometric dates within the interval, such as each timeseries to broaden the sample interval in steps U}Th dates or reservoir-corrected C-yr dates ad- until only a few samples remained within the LGM justed to the calendar scale using version 4 of the interval. For example, for the Byrd ice core dataset we CALIB software (Stuiver et al., 1998). resampled the 84 measurements within the Level 2 chro- E LGM Chronozone Level 2 18,000}24,000 cal-yr BP. nozone to 42, 21, 10, and 5 samples by iteratively culling Chronologic control based on two bracketing every second data point. radiometric dates of any kind within the interval 12 While each method may work well in speci"c circum- and 30 ka (i.e., within marine oxygen-isotope stage 2, stances, we found that the most stable estimate of the or by correlation of non-radiometric data to similar LGM values was the arithmetic mean, followed closely regional records that have been dated to match the by the trimmed mean. Values of individual data points level 1 protocol (for example, dO stratigraphy). nearest 21 ka, the minimum value, and the maximum E LGM Chronozone Level 3 18,000}24,000 cal-yr BP. value were all unacceptably sensitive to errors in dates or Chronologic control based on other stratigraphic con- data densities. The median and midrange estimates were straints (for example, a regional lithologic index such of intermediate stability in the data sets considered here. as %CaCO ) that are correlated to similar records For estimates of variability, the standard deviation dated elsewhere to match the level 2 protocol. was stable, even if only a few data values were available A variety of regional transient climate events may within a chronozone. The range of values in the chrono- occur, and it is unlikely that any time-slice study will be zone was not a stable measure of variability because it able to avoid all of them. Thus, the EPILOG group was highly dependent on the number of data points explored the possibility of data treatment protocols to included. A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 641

The di!erences in estimates of mean value and varia- nthic foraminifera. Other indices such as those based on bility between LGM chronozone levels 1 and 2 were not the shell size of certain species of planktonic foraminifera signi"cant, if the quality of dating was good; for example (BeH and Duplessy, 1976) have the potential to add in- the di!erence in LGM mean dO values between the formation in speci"c areas but are probably not uniquely two chronozone de"nitions are 0.02& at both W8709A- related to temperature or applicable in widespread areas. 13PC and in the GISP2 ice core, within analytical errors The EPILOG group discussed all methods in common of single measurements. In the Byrd ice core, the chrono- use, based on both published work and ongoing develop- zone de"nition produced slightly more error, about ments. 0.2&, but still small relative to the total glacial-interglacial range. In this ice core, the LGM level is centered 6.1.1. Faunal and yoral methods on a dO minimum (i.e., cold extreme), and is bounded Statistical prediction methods for estimating water- by relatively warm intervals. Dating errors of up to 1 kyr column properties from species assemblages are largely thus would induce small but signi"cant di!erences (up to empirical. The classical transfer function methods (Im- 0.4& dO) in mean LGM values of the Byrd core. brie and Kipp, 1971) resolve faunal assemblages in mod- Although larger than analytical precision of a single ern (core-top) samples, excluding redundant information analysis, these errors are small relative to the &10& via a factor analysis scheme. These assemblages are then contrast in glacial and interglacial values, and thus ac- calibrated in terms of sea-surface temperature or some ceptable for purposes of producing a global reconstruc- other property of the upper ocean using multiple regres- tion of the LGM. sion methods. The primary assumption is that modern Considering the tests done with these example spatial patterns of faunal abundance are controlled by datasets, we propose that EPILOG LGM data values be the same mechanisms that control changes through time presented as the arithmetic mean and standard deviation at any site. In theory, estimates of mean annual SST of all values within the chronozone, noting the level of using statistical transfer functions do not require the chronologic control included in the estimate. Although fauna or #ora to live at the sea surface, or in all seasons, we can imagine cases in which it would be appropriate to or even to be controlled mechanistically by temperature. cull clearly de"ned outliers from the LGM dataset, our The method requires, however, that whatever properties tests did not reveal any reliable method that could be actually control species distributions are linearly related applied to all data sets. We recommend that individual to SST. If this chain of relationships varies through time, investigators explore and document methods appropri- then estimates based on transfer functions may be biased. ate to their datasets. In all cases, analytical errors should Theory suggests that prediction of multiple environ- be presented as standard deviations, and for samples that mental properties is possible from rich multi-species contain less than three data values within the chronozone datasets (Imbrie and Kipp, 1971) but does not o!er the analytical error will be used as an estimate of the de"nitive guidance about which environmental proper- error for the mean value, but not as a measure of true ties are most important. Debate continues regarding the variability within the chronozone. potential for biases in estimates and the di$culty in uniquely assigning variations in the fauna to one or another environmental property, in part because of limits 6. New approaches to reconstructing LGM climate imposed by intercorrelation among environmental variables in the "eld (Watkins and Mix, 1998). Nevertheless, 6.1. The oceans most paleoceanographers accept that species assemblages can yield useful quantitative or semi-quantitative For the marine realm two methods are now commonly indices of upper-ocean temperature, as well as estimates applied for basin-wide mapping of surface-ocean of either biological productivity or pycnocline structure paleotemperatures; those based on species abundances of (noting that these two variables tend to be correlated in fossil plankton (using a variety of statistical techniques), the modern ocean). and those based on organic geochemistry (using di!erent Transfer functions based on the Imbrie and Kipp Y types of alkenone unsaturation indices, typically U). (1971) method were the primary source of information Oxygen isotope data based on planktonic foraminifera for the CLIMAP (1976, 1981) reconstructions of the have been used in support of temperature reconstruc- LGM oceans. Several faunal and #oral groups were tions, but rarely as a stand-alone temperature index considered, including foraminifera (prevalent in the At- because of the large (and in some cases unconstrained) lantic and Indian Oceans; McIntyre et al., 1976; Kipp, in#uences of ice-volume and local watermass e!ects. In- 1976), calcareous nannofossils (which received limited organic chemical methods that have been explored but use in the Atlantic and Paci"c Oceans; McIntyre, 1967; not yet widely applied include those based on trace metal Geitzenauer et al., 1976), Radiolarians (prevalent in the elemental ratios such as Sr/Ca, U/Ca, and Mg/Ca in Paci"c and parts of the Southern Oceans; Hays et al., aragonitic corals or in calcitic shells of planktic or be- 1976; Lozano and Hays, 1976; Moore et al., 1980), and 642 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 diatom-rich sediments (primarily in the Southern ents (the RAM method of Waelbroeck et al., 1998). In Oceans, Cooke and Hays, 1982). Concordance of transfer some cases, species percentage data is modi"ed with a log functions based on di!erent species groups strengthened transform, to minimize problems associated with species the CLIMAP reconstructions in some areas such as the dominance (Guiot, 1990; deVernal and Hillaire-Marcel, South Atlantic (Mol"no et al., 1982) but led to questions 2000). and a di$cult choice among con#icting estimates in Although MAT methods have been applied mostly to others, such as the tropical Paci"c (Moore et al., 1980). planktonic foraminiferal data, they have also been ex- CLIMAP (1981) chose to make seasonal reconstruc- tended to other planktonic microfossil assemblages tions of the earth's surface, based on calibration to Au- (Abelmann and Gowing, 1996; Abelmann et al., 1999; gust and February conditions. Questions arose regarding Brathauer and Abelmann, 1999; Crosta et al., 1998a). the statistical signi"cance of such seasonal estimates, Prell (1985) shows that the MAT and classical transfer- because modern calibration values from the two seasons function methods yield roughly similar results in the are highly correlated over large areas (Mix et al., 1986a, tropics. Further comparisons between methods using b; Pisias et al., 1997). CLIMAP chose these seasonal both living (sediment-trap) and fossil (core-top) assem- estimates in part because of limitations of computer blages suggest somewhat less bias in estimates for the models of the 1970s; model simulations typically were MAT than for transfer functions (Ortiz and Mix, 1997), performed as `perpetuala seasons rather than full sea- although a sparse data set for calibration such as exists in sonal cycles (Gates, 1976). Such limitations no longer the North Paci"c may yield imprecise or noisy estimates. apply. A potential problem for both transfer function and Recognition of possible problems and biases in statist- MAT methods occurs when a lack of suitable modern ical transfer functions, such as the estimate of seasonal analogs exists for a geologic sample. This is called the `no properties and preservation problems (for example, analoga problem. Such problems have been known for Parker and Berger, 1971; Berger and Gardner, 1975) led many years, for example in the eastern tropical Paci"c eventually to the development of alternate statistical and Atlantic Oceans (Moore et al., 1981; Mix et al., methods. Most widely used is the Modern Analog Tech- 1986a, b; Ravelo et al., 1990) and parts of the tropical nique (MAT), which attempts to match a geologic sample Indian Ocean (Hutson, 1977). The primary problem is from the past with a set of modern samples containing that faunal assemblages de"ned based on core-top sam- a similar fauna or #ora (Hutson, 1980; Prell, 1985). Envir- ples do not capture the full range of faunal variation that onmental estimates are then based on an average, or occurred during the ice ages. Mix and Morey (1996) and weighted average, of a number of best analog modern Mix et al. (1999b) address this issue, and develop a re- samples. The basic assumption of MAT is similar to that vised transfer-function strategy in which down-core sam- of classical transfer functions; i.e., that modern spatial ples, including those representing glacial conditions, are variability (in core-top samples) serves as a proxy for past used to de"ne more robust faunal assemblages. These temporal variability (down-core at a site). assemblages are then applied to core-top faunal counts, The MAT di!ers from transfer functions in providing and new transfer functions are calibrated in the normal an opportunity to reconstruct the full ensemble of envir- way. The down-core factor method improves the esti- onmental properties associated faunal or #oral popula- mates in areas without appropriate modern analogs, but tions. For example the MAT method might estimate does not remove all potential biases from the transfer a full seasonal cycle of upper-ocean temperature, pycnoc- function; for example the precision of the estimate ap- line depth, and seasonal productivity for a single sample, pears to be worse than the apparent precision of previous simply by averaging all the relevant atlas values in the estimates. A primary result of this procedure (Mix et al., analog sites. It is not clear to what extent all of these 1999b) is an estimate of LGM cooling in the eastern estimates are statistically signi"cant. Clearly only the equatorial Paci"c and tropical Atlantic oceans more ex- properties that truly control the populations would be tensive than that proposed by CLIMAP (1981), in better reliable, while other estimates would re#ect correlation of agreement with some geochemical data from oceanic properties within the calibration data set. Imbrie and sediments (for example, Hastings et al., 1998) and with Kipp (1971) note that for a sample set containing n stat- information on land (Hostetler and Mix, 1999). istically independent variables (such as faunal factors), at New mathematical procedures based on microfossil most n!1 environmental properties can be reconstruc- assemblages, such as arti"cial neural networks and re- ted. sponse surfaces (Malmgren and Nordlund, 1997) are also A variety of schemes have been explored to make an under development, with a goal of better isolating the average of the environmental properties in an array of e!ects of interdependent environmental variables such as best-analog samples, including a simple arithmetic aver- sea-surface temperature, mixed-layer depth, primary pro- age (Prell, 1985), weighting by geographic distance (the ductivity; all having the potential to a!ect the composi- SIMMAX method of P#aumann et al., 1996), or gridding tion of microfossil assemblages and induce biased and smoothing the core-top data along property gradi- estimates based on the other statistical methods. A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 643

Discussion of these methods by the EPILOG group 1cmka\, bioturbation may mix material of 10,000 suggested that all of the statistical methods are worth years or older into surface sediments (Peng et al., 1977). pursuing, with a goal of further assessing which method Because bioturbation moves material farther up the sedi- is most applicable in di!erent situations. For the ment column than it does downcore, LGM values are foraminifera, preparation of samples and taxonomic somewhat less sensitive to alternation of their mean by identi"cation is well established and large surface data mixing, as illustrated by benthic foraminiferal dO data sets are readily available, which make the microfossil (Zahn and Mix, 1991). techniques a very useful approach for global reassess- Following discussion of the EPILOG group, a consen- ment of LGM temperatures. Other fossil groups are less sus emerged that transfer function or MAT estimates of well developed, and additional e!orts will be needed to annual average and seasonal range (e.g., Mix et al., 1986a, develop these groups. Further work is needed to compare b; Pisias et al., 1997; Pisias and Mix, 1997) would be more the various statistical approaches. One important step appropriate than separate seasonal temperature esti- will be to make faunal datasets available publicly. Al- mates. Mean and range values are essentially indepen- though taxonomic concepts are reasonably well estab- dent of each other in many calibration data sets, and thus lished for the major fossil groups, the community would more likely to be robust as independent estimates. bene"t from regular intercalibration of taxonomies by Discussion continues on the accuracy and precision of sharing samples and data. the SST estimates based on di!erent statistical methods; Although most workers accept that faunal and #oral for example no consensus was reached about the e$cacy methods provide useful estimates of upper ocean proper- of geographical distance weighting between the site ties such as temperature and pycnocline depth, uncer- studied and best analogs (the SIMMAX method, tainties remain about the extent to which multiple P#aumann et al., 1996) and smoothing or gridding of environmental parameters can be extracted from a multi- species data sets along environmental gradients (the species data sets. For example, it is unclear whether RAM method, Waelbroeck et al., 1998) which both ap- faunas in the tropics are most appropriately used to pear to improve precision of MAT estimates. Neverthe- estimate temperature, pycnocline depth, or biological less, taking into account the errors of estimates (typically productivity, or whether the various in#uences result in $0.8}1.73C) calculated for the various statistical biased estimates (Mix, 1989; Ravelo et al., 1990; An- methods, results are fairly consistent and the choice of dreasen and Ravelo, 1997; Watkins and Mix, 1998; Beau- one method or another may depend on speci"c regional fort et al., 1999; Cayre et al., 1999; Cayre and Bard, 1999). issues. It is also unclear if faunal assemblages from the high latitudes make sense when applied to tropical samples, 6.1.2. Organic geochemical methods because the abundance of such polar species in equato- Ketone unsaturation ratios have gained a lot of atten- rial sediments of the LGM have no modern analog and tion in the last decade as a new tool for reconstructing are thus outside the range of statistical calibrations based past SST. These proxies were discussed at the EPILOG on surface sediments (Ravelo et al., 1990; Mix et al., 1999b). meeting, but "rm recommendations were deferred to Some species groups provide limited information at a meeting focused on alkenones, which occurred in fall of high latitudes. Foraminifera here have low species diver- 1999 at Woods Hole Oceanographic Institution (see the sity that limits the precision of estimates at low temper- forthcoming special issue of G, edited by T. Eglinton, J. atures. One possible solution, shown by U. P#aumann, is Hayes, M. Conte, and G. Eglinton; http://146.201.254.53/). to improve precision by counting a large number of Unsaturation ratios are thought to represent temper- specimens (perhaps thousands), so that relatively rare but atures during production of long-chain alkenones by important species may be identi"ed reliably. Other spe- haptophytes, a group of primary producers widespread cies groups may also resolve this problem. For example, in the ocean. Following early studies that identi"ed the dino#agellates are relatively diverse in the high latitudes, promise of alkenones as paleoceanographic proxies (e.g., and thus may provide useful estimates of multiple water Marlowe et al., 1984; Brassell et al., 1986; Prahl and column properties in these regions (deVernal et al., 1994). Wakeham, 1987), it has been shown that alkenone A potential problem for all methods that use core-top measurements in modern (core-top) sediments support Y # sediments to calibrate or con"rm proxy data in terms of the concept of using the U index ([C]/([C] temperature or other upper ocean data lies in the as- [C]) as a reasonable measure of SST (Herbert et al., sumption that core-top sediments re#ect modern condi- 1998; MuK ller et al., 1998). No signi"cant diagenetic e!ects Y tions. Many older sediment cores used in calibration on the U index have been detected (Prahl et al., 1989; studies have not been dated, and it remains unclear Madureira et al., 1995; Teece et al., 1998). Patterns of whether the core-top sediments are really modern. Bi- temperature change based on the index appear to behave oturbation mixes older material into surface sediments. in systematic ways through the last few hundred thou- This is a special problem in areas of low sedimentation sand years (Schneider et al., 1996, 1999; Bard et al., 1997). rates. For example, at sedimentation rates of less than The alkenones thus provide a powerful tool for 644 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 estimating LGM temperatures on a global scale, except concentrations yielded larger inter-laboratory variability Y for those regions where alkenone productivity and con- than the ratio index U. Such quanti"cation, although tents in sediments are very low, such as the centers of desirable, appears to have little e!ect on temperature subtropical gyres, and the polar oceans where temper- estimates in most cases. The most important aspects of atures are very low. the procedure are the quality of the gas chromatographic As with all geologic proxies, care must be taken to analysis and the skill of analyst, rather than the type of evaluate potential biases and errors on a case-by-case methodology. Chemical cleanup of organic extracts is basis. Second-order problems include the following: (1) not always required, but may be needed for some samples Observations of peak alkenone production at subsurface to obtain reliable alkenone data. A direct outcome of the Y depths suggest that the U index may not record true exercise will be the recommendation of analytical guide- Y sea-surface temperatures in all areas (Prahl et al., 1993; lines to carry out reliable U and alkenone abundance Okhouchi et al., 1999); (2) Observations of seasonal determinations (Rosell-MeleH et al., 2000). blooms in alkenone production suggest that the U in- A potential problem exists in some areas with very dex may not always record annual-mean SST (Prahl et cold or very warm waters at both ends of the alkenone al., 1993); (3) Kinetic e!ects of cellular growth, observed temperature calibration (e.g., Sikes and Volkman, 1993; Y to a!ect the calibration of the U index in some batch Sikes et al., 1997; Sonzogni et al., 1997). Here it becomes culture studies, suggest the potential for chemical or di$cult to properly quantify the minor peak on ecological biases (Epstein et al., 1998), although this was chromatograms (C and C for cold and warm not observed in steady-state cultures (Popp et al., 1998), temperatures, respectively) and analysis requires a high which raises questions about which type of culture ex- amount of alkenones preserved in sediments. A further periments adequately represent the real ocean; (4) The problem is deviation of the temperature calibration from Y "nding of di!erent relationships of U to temperature in a constant slope near the cold and warm ends of the di!erent species or strains of haptophyte algae in laborat- calibrated range. Such e!ects have been con"rmed in ory cultures leads to uncertainties in the choice of tem- cultures for some Haptophyte algae strains (Conte et al., perature calibrations applied to the geologic record in 1998). However, in general the error of SST estimates is some areas (Conte et al., 1998); (5) Erosion and redeposi- similar to that in microfossil assemblage techniques and tion at the sea #oor suggest that alkenone signatures in ranges between 1 and 1.53C, depending on whether a re- some areas such as sediment drifts could be con- gional basin-wide or the global surface sediment calib- taminated with material that is older or from another ration is used (MuK ller et al., 1998). geographic area (Weaver et al., 1999; Benthien and MuK l- Alkenone records are often in agreement with other ler, 2000); and (6) Bioturbation in the sediment column, records in the low- and mid-latitude bands. For the which appears to be strongest for "ne particles that carry mid-latitude sites studied by Prahl et al. (1995) and Pi- alkenones, may slightly shift and di!erentially smooth chon et al. (1998) there is a good agreement between the stratigraphic record of alkenone indices relative to alkenone results and temperature estimates based on other proxy data carried on larger particles (Bard, sub- radiolarians and diatoms, respectively. mitted 2000). Some relatively large di!erences have been noted in The EPILOG meeting included a discussion of an several instances. For example, to reconcile transfer func- international laboratory intercomparison experiment, tion, alkenone, and dO records measured in a core coordinated by the European TEMPUS project located o! Mauritania, Chapman et al. (1996) invoked (Rosell-MeleH et al., 2000). In this coordinated e!ort 26 large changes of the seasonality of foraminiferal growth laboratories worldwide participated in an anonymous and coccolithophorid growth. However, further work on intercomparison exercise and analyzed (doing several Mg/Ca in the same core (Elder"eld and Ganssen, 2000) replicates) "ve unmarked sediment samples and three suggests a limited cooling of about 2}33C during the alkenone standard mixtures with a wide range of LGM, in agreement with alkenone and dO records. In Y U values and concentration of alkenones. The experi- an Indian Ocean core, temperature changes based on ment evaluated whether there are di!erences amongst the foraminiferal species systematically lead those based on Y Y laboratories results for U and alkenone absolute con- U (Cayre and Bard, 1999), although the total range of centration that are signi"cant for climate reconstruc- temperatures is similar in both proxies. Y tions, and if so to what extent this relates to the analytical In the subtropical Atlantic, regional U estimates Y methodology employed. In relation toU, it is remark- exhibit less glacial-to-interglacial change than estimates able that despite being the "rst large exercise of its kind, based on Sr/Ca or U/Ca in corals (Guilderson et al., most laboratories have provided data that are signi"- 1994; Min et al., 1995). Y cantly similar (95% con"dence). However, there is room In the equatorial Atlantic, Mg/Ca and U records to improve the methodology in some instances, and not both suggest slight cooling (2}33C) at the LGM, al- all laboratories are capable of analyzing all the samples though the patterns of change through time are some- with the same pro"ciency. Quanti"cation of alkenone what di!erent, and Mg/Ca temperature estimates A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 645

(measured in the foraminiferal species G. sacculifer) are 6.1.3. Trace metal methods Y cooler than U estimates in the same samples by 0}33C The search for new and independent paleothermomet- (NuK rnberg et al., 2000; Schneider et al., 1996). The ric methods has been accelerated recently by the appar- authors attribute these di!erences to habitat preferences ent discrepancy between the classical CLIMAP of G. sacculifer, which in this area is most prevalent reconstruction and the Sr/Ca ratios measured in corals during cool upwelling events of the austral fall and win- from Barbados (Guilderson et al., 1994). This widely ter. In contrast, the alkenone producers in this region are publicized study is based on a single site and is often used thought (based on core-top correlations) to live in near- as evidence that CLIMAP's foraminifera-based SSTs surface waters throughout the year (MuK ller et al., 1998). were much too warm in the intertropical zone and that Transfer-function and modern-analog estimates of sea- a severe cooling of about 53Ca!ected the warm pools sonal SST based on foraminiferal species distributions during the LGM. Comparison of this result to other (Wefer et al., 1996) suggest larger temperature changes nearby data based on many techniques, however, suggest Y than U or Mg/Ca in the same core (NuK rnberg et al., that cooling based on the Sr/Ca estimate is at the extreme 2000). This careful multi-proxy comparison highlights of all methods. Crowley (2000) points out that cooling of the di$culty in judging which estimate is the most appro- 53C in the tropics would fall below the survival temper- priate measure of mean annual sea-surface temperature. ature of most corals and would likely have restricted the High-latitude regions also display con#icts between LGM coral distribution to perhaps 5% of its present indices. For example, in the Southern Ocean, alkenone- range, an outcome for which there is no evidence. based temperatures are warmer by about 23C than those For corals, several authors have shown that besides derived from diatom species abundances (Pichon et al., temperature the recorded Sr/Ca can be a!ected by biolo- 1998). For a North Atlantic site at 563N, Sikes and gical control on Sr and Ca uptake (Ip and Krishnaveni, Keigwin (1994) reported a similar amplitude and timing 1991; Ip and Lim, 1991; de Villiers et al., 1995), growth of deglacial warming with alkenone and foraminiferal rate e!ects (Weber, 1973; de Villiers et al., 1994), intra- or species data. However, the SST estimates derived from inter-species dependence (Weber, 1973; Alibert and alkenones are about 43C higher than those based on McCulloch, 1997; Boiseau et al., 1997; Cardinal et al., foraminifera. Weaver et al. (1999) observed a similar 2000), skeleton heterogeneity (Allison, 1996; Hart and pattern in a core located at 593N. These authors suggest Cohen, 1996; Greegor et al., 1997), and bias in calibration that both records could be reconciled by using a di!erent due to distortions of seasonal signal (Barnes et al., 1995; calibration that assumes that alkenone data are represen- Cardinal et al., 2000). The preservation of the Sr/Ca tative of summer growth conditions rather than mean signal may not be always ideal because of diagenetic annual growth conditions. dissolution (Amiel et al., 1973; Cross and Cross, 1983; One common problem encountered in sediments from Yoshioka et al., 1986). The interpretation of the measured the northern North Atlantic (Sikes and Keigwin, 1996; Sr/Ca may be complex because of changes of seawater Weaver et al., 1999) is the rather low concentration of Sr/Ca (deVilliers et al., 1994; Shen et al., 1996; Stoll and alkenones, which complicate the standard GC measure- Schrag, 1998; de Villiers, 1999). Recently, Martin et al. ments. For example, Weaver et al. (1999) had to measure (2000) proposed a new reconstruction of the oceanic alkenones by GC-MS-CI in sediments with concentra- Sr/Ca ratio that exhibits large secular changes with maxi- tions ranging between 0.1 and 10 ng g\. Such North ma on the order of 3}4% in phase with glacial maxima. Atlantic sediments are probably more susceptible to con- The immediate conclusion derived by these authors is tamination by reworking. Indeed, Weaver et al. (1999) that glacial SST changes derived from Sr/Ca in corals observed that in the glacial section of their core, `the have been overestimated by at least a factor 2. in-situ #ora becomes very rare and the number of rewor- U/Ca and Mg/Ca are also potentially useful indices of ked Cretaceous and Paleogene specimens increases dra- temperature in corals. However, U/Ca could be a!ected maticallya. by changes of the carbonate ion concentration (Min et The EPILOG group consensus is that alkenone indi- al., 1995; Lea et al., 2000) and/or the salinity (Shen and ces o!er great potential for widespread reconstructions of Dunbar, 1995). The interpretation of the measured U/Ca upper-ocean temperatures at the LGM. More work is may be complicated because of intra- or inter-species needed to sort out the speci"c biases (especially those of dependence (Min et al., 1995; Cardinal et al., 2000). Sim- seasonal or subsurface production), but these e!orts ilarly, the interpretation of the measured Mg/Ca in corals should not impede the use of alkenones as part of may be complicated by growth-rate e!ects (Swart, 1981). a multi-proxy strategy for reconstructing properties of Several authors have observed unidenti"ed scatter the ice-age oceans. Alkenones may be especially useful in and/or discrepancies between calibrations (Schrag, 1999; areas of relatively high biological production (such as Sinclair et al., 1998), and heterogeneous records within near continental margins) and where other fossil groups coral skeletons (Allison, 1996). In addition, the diagenetic are poorly preserved (such as in deep water where calcite dissolution is identi"ed as a potential bias (Amiel et al., tends to dissolve). 1973; Cross and Cross, 1983). For both U and Mg in 646 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 corals, more work remains to be done to assess these based on radiocarbon ages as 18,000$2000 C-yr BP. methods. This de"nition is su$ciently close to the EPILOG calen- Determinations of Mg/Ca ratios in foraminiferal shells dar age de"nition of LGM Chronozone Level provide a promising new tool for estimating not only 1 (19,000}23,000 cal-yr BP), for comparison to new data sea-surface temperature (Rosenthal et al., 2000) but also from the ocean. A di!erence from the EPILOG de"ni- intermediate or deep-water temperatures when applied tion, however, is that the Farrera et al. (1999) included at also to planktonic deep dwellers or benthic foraminifera each site the single data point closest in age to 18,000 and ostracods (Dwyer et al., 1993). One advantage of this C-yr BP, rather than an average of values within the method is its combination with measurements of oxygen LGM time interval. isotopes at the same material, which o!ers the potential Data sources on land include pollen, plant macrofos- to estimate paleo-dO of seawater. The combination of sils, lake levels, noble gases in groundwater, and dO isotopes and elemental ratios then may help to constrain from speleothems. The large international group contri- past salinity variations. Knowledge about paleo dO buting to the Farrera et al. (1999) study assessed which water}salinity relationships in di!erent regions is still climatic variables were best inferred from their data, and needed for precise salinity estimates. Application of cli- concluded that estimates of mean temperature of the mate models that incorporate isotope tracers may help to coldest month, mean annual temperature, plant-avail- estimate the paleo dO of water vapor and river runo! able moisture, and runo! (i.e., precipitation less evapor- (Rohling and Bigg, 1998; Schmidt, 1999; Delaygue et al., ation) were most appropriate. 2000). E!orts are now underway to build a global In this compilation, mean annual temperatures in database of dO and salinity data, to improve the basis tropical lowlands cooled by 2.5}33C at the LGM, only for such modeling e!orts (G. Schmidt, pers. comm. 2000; about 1.53C more than tropical ocean cooling inferred by http://www.giss nasa.gov/data/018data). CLIMAP (1981). The results varied by area, however, Besides temperature, the recorded Mg/Ca can be a!ec- with LGM cooling estimated to be 5}63C in Central ted by cleaning procedure, pH, and salinity (Lea et al., America and northern South America, 13CinNew 1999). Several authors have observed complications such Guinea and the Paci"c Islands, and 2}33C in Africa and as intra- or inter-species dependence (NuK rnberg et al., Southeast Asia. High-altitude sites suggest much greater 1996), unidenti"ed scatter or bias in the calibration, i.e., average LGM cooling, however. For example, sites near cultures vs. core tops (NuK rnberg et al., 1996; Mashiotta et 3000 m elevation cooled about 63C in the Andes al., 1999), shell heterogeneities due to secondary growth (Thompson et al., 1995), New Guinea and Africa (Farrera of gametogenic calcite (Brown and Elder"eld, 1996; NuK r- et al., 1999). This suggests signi"cantly steeper vertical nberg et al., 1996). The in#uence of dissolution has been lapse rates (temperature change with altitude) during the noted (Rosenthal and Boyle, 1993; Brown and Elder"eld, LGM, accompanied by a weaker hydrologic cycle. Farr- 1996; Rosenthal et al., 2000). era et al. (1999) suggest that terrestrial lapse rates (de"ned Although the elements are simple to measure the as the vertical temperature gradient on the earth surface) method requires proper cleaning techniques of the cal- may have diverged signi"cantly from free-air lapse rates. citic shells, which should be standardized and calibrated As noted above, modeling studies of the Paleoclimate between di!erent labs. Since this method is fairly new no Modeling Intercomparison Project (PMIP), summarized global or basin-wide surface-sediment calibration against by Pinot et al. (1999), concur with the earlier study of atlas values exist for planktonic foraminifer Mg/Ca ra- Rind and Peteet (1985) that it is di$cult to reconcile the tios. Culture experiments have been carried out so far CLIMAP (1981) ocean reconstruction with terrestrial only for di!erent surface dwellers, e.g., G. ruber, G. sac- evidence from the tropics. Atmospheric lapse rates varied culifer, (NuK rnberg et al., 1996), and for G. bulloides slightly in the models, but not as much as implied by the (Lea et al., 1999). Caution is also needed because of data of Farrera et al. (1999) in some areas. Experiments potential bias due to calcite dissolution and ecological with computed sea-surface temperatures in the models factors, e.g., seasonality and depth of calci"cation, as well with interactive mixed-layer oceans suggested that tropi- as mineralogical heterogeneity in the shells (gametogenic cal sea-surface cooling was slightly greater than that of calcite). These factors have so far kept this method from CLIMAP (1981), and may vary regionally. general use, but it should be considered an innovative That lapse rates do not change substantially in atmo- technique that deserves further attention. spheric GCM's does not prove their stability, however, as these models parameterize (i.e., simplify) sub-grid scale 6.2. The continents vertical convection in the atmosphere. In a detailed, one-dimensional radiative-convective model, Betts and Building on earlier global compilations of land cli- Ridgeway (1992) found signi"cant changes in free-air mates by CLIMAP (Peterson et al., 1979) and COHMAP lapse rates associated with surface winds (steeper lapse (1988), Farrera et al. (1999) surveyed land climates of the rate with lower winds), or poleward heat transport tropics during the LGM. These studies de"ne the LGM (steeper lapse rate with greater heat transport). They A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 647 concluded that observed advances in tropical glaciers are identify key opportunities to "ll gaps in our current possible even with CLIMAP's (1981) mean tropical cool- knowledge of the LGM. ing of 1.43C, but would be more plausible with slight additional cooling of the sea surface totaling 23C or more 7.1. High latitudes averaged over the tropics. Crowley (2000) also assessed the e!ect of CLIMAP In the North Atlantic region, several new LGM SST sea-surface temperature on continental climates. He ar- syntheses are currently in progress using transfer func- gued that apparent mis"ts with tropical snowline data tion or modern analog techniques applied to assemblages may not be as large as previously thought, and that of foraminifera and dino#agellates, as well as with or- relatively small changes in free-air lapse rate, coupled to ganic geochemical methods. Preliminary results agree modest additional sea surface cooling of about 13C, that ice-free surface waters existed during summer season would be enough to explain most of the terrestrial data. in the eastern North Atlantic and Norwegian Sea, per- Hostetler and Mix (1999) "nd that regional cooling in the haps as far north as 703N, but di!erences in estimates of tropical Atlantic and eastern Paci"cissu$cient to ex- sea-surface temperature have raised controversy. plain high altitude glaciers and changes in lake level in In the Nordic seas, LGM temperature estimates based tropical South America and Africa, and especially note on foraminiferal modern analogs (SIMMAX method of that parts of the Andes were wetter as well as cooler, in P#aumann et al., 1996) suggest relatively cold sea-surface contrast to tropical aridity in the lowlands. Crowley conditions, near freezing in winter and between 2 and (2000) and Crowley and Baum (1997) point to further 53C in summer (Sarnthein et al., 1995; Weinelt et al., work in understanding the roles of air}sea interaction, 1996). In contrast, temperature estimates based on dino- tropical convection, changing vegetation, and the e!ects #agellates (deVernal, 1993, 2000) and on the alkenone Y of atmospheric moisture in setting model sensitivity. Re- index U (Rosell-MeleH and Comes, 1999), both suggest solving the e!ects of these variables in controlling LGM seasonally warm sea-surface temperatures, perhaps as climates are important because all have implications for high as 133C * similar to or in some cases even warmer prediction of future climate changes. than modern values. This is surprising, given "ndings Both the modeling studies, and the data compilation of from ice cores in Greenland suggesting even greater cool- Farrera et al. (1999) support the concept of LGM aridity ing that previously estimated (Cu!ey et al., 1995; Johnsen in many areas of the tropical lowlands, although this et al., 1995). Alkenone SST estimates may be questioned concept has been challenged in some areas (Colinvaux due to very low production of alkenone-forming organ- et al., 2000). Other challenges for continental LGM re- isms in cold water and the possibility of contamination constructions discussed during the EPILOG meeting with allochthonous older material. In addition, the al- included (1) the di$culty in dating many land records, (2) kenone estimates for very low temperatures may be bi- "nding adequate methods for inferring regional patterns ased by extraordinary amounts of the C ketone from sparse local records, and (3) the need for better reaching more than 10% of total C alkenones, a fea- methods to reconstruct of past lapse rates, including the ture which may be related to reduced salinities (Rosell- di!erences between terrestrial lapse rates inferred by MeleH , 1998). Given these uncertainties, Rosell-MeleH and Farrera et al. (1999) and free-air lapse rates that are Comes (1999) discount the traditional alkenone temper- calculated in GCMs. Land proxies such as pollen or ature index, but develop an alternate index that yields eolian dust in marine sediment cores may help synchro- LGM temperatures associated with summer bloom con- nize climate information from land and sea. Analysis of ditions in the Nordic seas near 63C. regional climate models o!ers the potential to under- The apparent disagreement between the LGM temper- stand linkages between large-scale climate patterns in ature estimates based on foraminifera and those based on global GCMs and geologic data in speci"c locations. As dino#agellates in the Nordic seas remains. To resolve this with all geologic proxies, further work on intercalibra- disagreement better knowledge is needed about possible tion and coordination of methodologies will be needed di!erences in the depth habitats and the living seasons of for detailed global-scale reconstructions. foraminifera and dino#agellates in the Nordic seas. Per- haps the dino#agellates respond to environmental parameters other than SST. Perhaps they record temperature, 7. Toward a regional synthesis of the LGM but under no-analog conditions of a water column strongly strati"ed in summer by fresh-water inputs from Here we assess progress toward a new reconstruction melting glaciers. Perhaps the foraminifera record lack the of the Earth's surface during the LGM interval in various resolution to resolve temperature due to dominance by regions. Although a full synthesis is beyond our reach, we a single species in cold environments, or perhaps they highlight points of agreement, as well as unresolved con- re#ect subsurface conditions (Kohfeld et al., 1996). Dis- troversies. Our primary purpose is to outline strategies cussion at the EPILOG meeting highlighted the import- that will lead to resolution of remaining con#icts, and to ance of resolving the apparent di!erences between 648 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 indices by better understanding the physical and biolo- an e!ective barrier to exchange of heat, water vapor and gical controls on each biotic group. gases (such as CO) between the ocean and atmosphere The North Paci"c remains a poorly known ocean; due (e.g., Ramstein and Joussaume, 1995; Stephens and Keel- to much more limited sampling than has occurred in ing, 2000). other ocean basins and di$culties in de"ning the LGM Reconstruction of the Southern Ocean remain limited chronozone due to poor preservation of carbonate. This by the di$culty in linking local stratigraphic records into region is particularly important as an upwind source of a global framework, due largely to the absence of cal- heat and moisture to North American ice sheets (Peteet careous fossils for stable isotope stratigraphy or et al., 1997), and will require focused attention. Here, radiocarbon dating in many southern high-latitude sedi- CLIMAP estimates of LGM temperatures were uncer- ments. In some cases a lack of sediment cores (especially tain, and subject to some arti"cial corrections to adjust in the Paci"c sector) also limits progress. Development of for perceived transfer function bias (Moore et al., 1980). additional stratigraphic tools, such as isotope and/or Since CLIMAP (1981), it has become clear that substan- metal indices in siliceous sediments, and further "eld tial changes in upper ocean structure of the NW Paci"c studies, may be helpful in this region. were associated with increased intermediate water formation in the Sea of Okhotsk (Keigwin, 1998). However, the 7.2. Low to mid-latitudes Gulf of Alaska appears to have been more strati"ed at the LGM than at present (Zahn et al., 1991). This strati"- For the low and mid latitudes new LGM SST compila- cation, along with strong cooling, appears to have yiel- tions are developing based on a variety of faunal transfer ded seasonal sea ice cover in the NE Paci"c during functions and other statistical approaches noted above. LGM, based on dynocyst assemblages (deVernal and A consensus appears to be emerging that sea-surface Pedersen, 1997). temperatures in the tropics were slightly but not severely In the northern source areas of the California Current, cooler than those of CLIMAP (1981), and that temper- Ortiz et al. (1997) used foraminiferal and isotopic data to ature reduction was probably not uniform within the infer LGM cooling of about 43C, driven by small south- tropics (Broecker, 1986; Mix et al., 1986a, b; Broccoli and ward translation of the subpolar front. Such cooling is Marciniak, 1996; Pisias and Mix, 1997; Rosell-MeleH et al., consistent with other estimates based on radiolarians 1998; Bard, 1999; Mix et al., 1999a, b; Broccoli, 2000; (Sabin and Pisias, 1996), and alkenone indices (Prahl Crowley, 2000). et al., 1995; Doose et al., 1997). Con#icts appear o! CLIMAP's (1981) data coverage was sparse in the Southern California, however, where alkenone and subtropics, especially in the Paci"c. For example, recent radiolarian indices suggest relatively small LGM cooling data suggest signi"cant LGM cooling near Hawaii (Lee of about 23C (Herbert et al., 1995; Sabin and Pisias, and Slowey, 1999), where CLIMAP suggested either no 1996), whereas foraminifera and stable isotopes suggest change or seasonal warming. Broccoli and Marciniak larger cooling of '53C in the Santa Barbara basin (1996) note that many apparent con#icts between data (Hendy and Kennett, 1999). The EPILOG group noted and models occurred where CLIMAP interpolated re- that the North Paci"c will require much more work to sults between sparse or uncertain data points. Signi"cant de"ne patterns and mechanisms of change during the con#icts between various proxy methods remain, espe- LGM, and that apparent disagreements between proxies cially between Sr/Ca data from corals and other data in in the region may in some cases re#ect true spatial varia- the warm pool areas of the Atlantic and Paci"c (Crowley, bility of this large and dynamic region. 2000), and such con#icts need to be resolved. Uncertain- Progress in the Southern Ocean has come from better ties also remain about the spatial pattern and gradients understanding of the ecology of diatoms and radiolarians of low-latitude temperature changes, which in climate at all latitudes (e.g., Zielinski and Gersonde, 1997; models have proven important in changing the distribu- Welling and Pisias, 1998; Abelmann et al., 1999). Such tion of precipitation (Hostetler and Mix, 1999; Yin and information applied to down-core records has yielded Battisti, 2000, in press). better knowledge of open ocean temperatures (e.g., CLIMAP (1981) reconstructed relatively little change Pichon et al., 1992; Brathauer and Abelmann, 1999), as in SST, about 13C, in the western Paci"c warm pool. well as revisions to reconstructions of maximum extent More recent studies based on faunal transfer functions or and seasonal meltback (or fractional coverage of) sea ice Modern Analog methods have generally con"rmed that (Burckle et al., 1982; Burckle, 1984, Burckle and Mor- "nding, but have shrunk the region of stability. For tlock, 1998; Crosta et al., 1998a, b; Armand, 2000). Infer- example, foraminiferal estimates support stability of the red sea-ice distributions, especially in southern summer, warm pool, but also imply signi"cant cooling o! eastern are substantially di!erent than those of CLIMAP (1981), Australia and in the marginal basins o! southeast Asia, and this deserves further attention due to the importance both areas where CLIMAP had little data (Anderson et of sea ice as a high albedo surface, as a thermal insulator al., 1989, Miao et al., 1994; Thunell et al., 1994). Barrows decoupling ocean and atmospheric temperatures, and as et al. (2000) con"rm this result, but also note that the A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 649 coolest time in this area may not be the LGM. These global system to change, and for evaluating climate mod- "ndings of small SST changes within the oceanic warm els that will be used to predict future changes. Here we Y pool are supported by U data (Okhouchi et al., 1994), recommend key protocols and strategies so that the glo- although Sr/Ca in corals from the region suggests sub- bal community of earth scientists can join together in stantial temperature changes within the Holocene (Beck creating this new synthesis, which has come to be known et al., 1997), in apparent con#ict with regional estimates as EPILOG (Environmental Processes of the Ice age: based on foraminifera. Land, Oceans, Glaciers). Speci"c "ndings and recom- The extent to which upper ocean processes manifested mendations arising from the "rst open meeting of EPI- in pycnocline depth and/or density contrast (Andreasen LOG, held in Delmenhorst, Germany in May 1999, and Ravelo, 1997; Mulitza et al., 1997; Watkins and Mix, include the following: 1998; Ravelo and Andreasen, 1999; Wol! et al., 1999) can The EPILOG de"nition of the LGM is chronologic, be assessed remains uncertain. Progress here o!ers the based on ages 19}23 cal-kyr BP (Chronozone level 1), or potential to reconstruct the strength of ocean currents 18}24 cal-kyr BP (Chronozone level 2). A level 3 chrono- based on paleogeostrophy calculations. zone is de"ned with the same ages as level 2, but with less It was recommended during discussion of the various certainty about dating. We recommend conversion of regional studies of low- and mid-latitudes that annual radiocarbon dates to calendar ages using the CALIB-4 mean SST and seasonal contrast should be estimated correction scheme and that all chronologic data be pre- rather than seasonal temperatures. This strategy accom- sented in both uncorrected and corrected forms. plishes much the same thing as CLIMAP's (1981) esti- It is desirable to know both the mean value and varia- mate of seasonal temperatures, but helps to maintain bility of paleoenvironmental estimates within the LGM statistical independence of the calibration data sets. chronozone. We evaluated a variety of statistical treat- If the di!erences between the various proxies of tem- ments to assess these properties while excluding climatic perature can be resolved and understood, additional in- transients or outliers, and found that in most cases the sights into climate systems may come from combinations arithmetic average and standard deviation of available of proxies. For example, the combination of oxygen iso- data points within the chronozone at each site is the most tope measurements on foraminifera with independent reliable estimate of each property. Our tests were not estimates of temperature o!ers potential insight into cha- exhaustive, however. We recommend that individual nges in sea-surface salinity related to regional distribu- scientists evaluate their data sets to "nd the best way of tions of evaporation and precipitation, which is representing their data in a community reconstruction. important in tropical climate systems. An ultimate goal is to understand the transitions into and Further progress on all proxies will come from coor- out of the LGM, as well and the mean state and variabil- dinated study of the same samples with multiple proxies. ity within the chronozone. The EPILOG group encourages public distribution of A major change since the completion of the CLIMAP datasets from both modern and LGM samples via estab- project is the rise of multi-proxy approaches, with new lished databanks (for example, the National Geologic statistical methods applied to faunal and #oral data, and Data Center, NGDC, in the US, and the PANGEA new geochemical indices. With this new work, we note databank in Europe). For proxies that require calib- points of emerging consensus, such as that tropical sea- ration with modern atlas data, we encourage all calib- surface temperatures cooled slightly (but not dramati- rations to be done with one data product of uniform cally) more than suggested by CLIMAP (1981), as well as quality. Use of the World Ocean Atlas, WOA98 points of controversy, such as seasonal temperatures and (National Oceanographic Data Center, 1999) for sea-ice cover in the high latitudes of both hemispheres. purposes of proxy calibration is recommended, (http:// Resolution of these and other con#icts will likely yield www.nodc.noaa.gov/OC5/woa98.html). deep insights into the proxies and their underlying systems, and into the processes driving global change. To facilitate intercomparison between proxies, and between 8. Conclusions and general recommendations various workers using similar proxies, EPILOG recom- mends (1) intercalibration experiments, following the Much progress in understanding climates of the LGM model of TEMPUS (Rosell-MeleH et al., 1998), (2) archival has occurred in the &20 years since the end of the of available datasets in public databases, (3) e!orts to CLIMAP project. New geochemical methods and new share samples to the extent possible, so that comparisons statistical approaches to reconstruction of past environ- among datasets can be made with rigor, and (4) con- ments reveal signi"cant modi"cations are needed in the tinued e!orts to obtain high-quality materials through CLIMAP (1981) reconstruction of the surface of the ice- "eld programs. For calibrations involving use of atlas age Earth, and a new synthesis is warranted. Such a syn- data, we recommend use of a uniform dataset * for thesis will provide an important benchmark for under- example in the ocean; the World Ocean Atlas 98 provides standing the processes that set the sensitivity of the a suitable product. 650 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

It is clear that geographic coverage of LGM sample reviews, we thank T. Crowley and T. Herbert, and P. sets is insu$cient for meaningful global reconstructions Clark. For additional discussions we thank N. Pisias. For in the ocean or on land. Gaps in sample coverage are "nancial and logistical support, we thank the Hanse particularly acute in the Paci"c, Antarctic, and Arctic Institute for Advance Study, the University of Bremen, Oceans, and on land areas outside of North America and IMAGES, IGBP/PAGES, the European Science Foun- Europe. However, attention to climatically important dation (grants ENV4-CT97-0564, FMRX-CT96-0046, gradients in all regions is warranted. and ENV4-CT97-0659, to EB) and the US National Although the "rst EPILOG meeting concentrated on Science Foundation (grants OCE9632029, OCE9730376, reconstruction of sea-surface temperature and related and OCE9818237 to ACM). Sediment core archival at upper ocean properties as well as tropical land temper- Oregon State University is supported by NSF grant atures, a full understanding of the LGM will require OCE9712024. a much broader perspective, including detailed reconstructions of glaciers and ice sheets, the land surface and its vegetation, a number of sea-surface properties including seasonal and annual temperature, salinity, productiv- References ity, and upper ocean structure, and changing character of the deep-sea and thermohaline circulation. Rather than Abelmann, A., Brathauer, U., Gersonde, R., Sieger, R., Zielinski, U., making seasonal reconstructions of the sea or land sur- 1999. Radiolarian-based transfer function for the estimation of sea surface temperatures in the Southern Ocean (Atlantic Sector). face we recommend estimation of annual means and Paleoceanography 14, 410}421. season ranges of ocean properties. This accomplishes Abelmann, A., Gowing, M.M., 1996. Spatial distribution pattern of much the same thing as seasonal estimates, but helps to living polycistine radiolarian taxa: baseline study for paleoenviron- maximize statistical independence of real-world calib- mental reconstructions in the Southern Ocean (Atlantic Sector). ration data sets based on spatially distributed modern Marine Micropaleontology 30, 3}328. Alibert, C., McCulloch, M.T., 1997. Strontium/calcium ratios in mod- samples. ern Porites corals from the Great Barrier Reef as a proxy for sea Desirable products for a new LGM reconstruction surface temperature: calibration of the thermometer and monitoring include the mean state of the LGM, variability within the of ENSO. Paleoceanography 12, 345}363. LGM interval, and the transitions into and out of the Alley, R.B., Meese, D.A., Shuman, C.A., Gow, A.J., Taylor, K.C., LGM state. Creating a synthesis of such breadth is too Grootes, P.M., White, J.W.C., Ram, M., Waddington, E.D., Mayewski, P.A., Zielinski, G.A., 1993. Abrupt increase in Greenland large a task an individual or even a small group of snow accumulation at the end of the Younger Dryas event. Nature scientists. Instead, we envision an open e!ort in which 362, 527}529. the global community of earth scientists contributes to Alley, R.B., Shuman, C.A., Meese, D.A., Gow, A.J., Taylor, K.C., Cu!ey, a group synthesis. To facilitate this e!ort, we encourage K.M., Fitzpatrick, J.J., Grootes, P.M., Zielinski, G.A., Ram, M., the organization of regular topical meetings spanning Spinelli, G., Elder, B., 1997. Visual-stratigraphic dating of the GISP2 ice core: basis, reproducibility, and application. Journal of several years. The next EPILOG meeting, to be held in Geophysical Research 102, 26,367}26,381. October 2000 (facilitated by P. Clark, E. Bard, A. Mix), Allison, N., 1996. Geochemical anomalies in coral skeletons and their will focus on the reconstruction of the global ice sheet possible implications for palaeoenvironmental analyses. Marine system. Chemistry 55, 367}379. Amiel, A.J., Miller, D.S., Friedman, G.M., 1973. Incorporation of uranium in modern corals. Sedimentology 20, 47}64. Anderson, D., Prell, W., Barratt, N., 1989. Estimates of sea-surface Acknowledgements temperature in the Coral Sea at the Last Glacial Maximum. Paleoceanography 4, 615}627. For scienti"c contributions and spirited discussions, Andreasen, D.J., Ravelo, A.C., 1997. Tropical Paci"c Ocean thermoc- we thank all the participants in the EPILOG "rst open line depth reconstructions for the Last Glacial Maximum. Paleoceanography 12, 395}413. meeting; D. Anderson, T. Barrows, H. Behling, P. Clark, Armand, L.K., 2000. An ocean of ice * advances in the estimation of E. Cortijo, P. De Deckker, S. De Rijk, A. DeVernal, M. past sea ice in the Southern Ocean. GSA Today 10/3, 2}7. Diepenbroek, S. Drachenberg, G. Eglinton, H. Elder"eld, Bard, E., 1999. Ice age temperatures and geochemistry. Science 84, K.-C. Emeis, A. FluK gge, B. Grieger, J. Grimalt, H. Grobe, 1133}1134. J. Guiot, S. Harrison, H. Hedges, T. Herbert, H. Bard, E., 2000. Paleoceanographic implications of the di!erence in deep-sea sediment mixing between large and "ne particles. Hooghiemstra, S. Hostetler, E. Ivanova, E. Jansen, D. Paleoceanography, submitted. Jolly, S. Joussaume, S. Juggins, K. Lambeck, D. Lea, A. Bard, E., Arnold, M., Duplessy, J.-C, 1991. Reconciling the sea level Mazaud, I. Mehser, S. Mulitza, P. MuK ller, H. Niebler, J. record of the last deglaciation with the dO spectra from deep-sea PaK tzold, U. P#aumann, J.-J. Pichon, C. Porthun, W. cores. Quaternary Proceedings 1, 67}73. Prell, R. Ramseier, A.C. Ravelo, A. Rosell-MeleH ,Y. Bard, E., Arnold, M., Hamelin, B., Tisnerat-Laborde, N., Cabioch, G., 1998. Radiocarbon calibration by means of mass spectrometric Rosenthal, M. Sarnthein, M.-A. Sicre, R. Spielhagen, M. Th/U and C ages of corals. An updated database including Trend-Staid, J.-L. Turon, E. Vogelsang, C. Waelbroeck, samples from Barbados, Mururoa and Tahiti. Radiocarbon 40, G. Wefer, M. Weinelt, and U. Zielinski. For thoughtful 1085}1092. A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 651

Bard, E., Arnold, M., Mangerud, J., Paterne, M., Labeyrie, L., Duprat, Bond, G., et al., 1992. Evidence for massive discharges of icebergs into J., MeH lie`res, M.-A., S+nstegaard, E., Duplessy, J.C., 1994. The North the North Atlantic Ocean during the last glacial period. Nature 360, Atlantic atmosphere-sea surface C gradient during the Younger 245}249. Dryas climate event. Earth and Planetary Science Letters 126, Brassell, S.C., Brereton, R.G., Eglinton, G., Grimalt, J., Liegezeit, G., 275}287. Marlow, I.T., P#aumann, U., Sarnthein, M., 1986. Paleoclimatic Bard, E., Hamelin, B., Arnold, M., Montaggioni, L., Cabioch, G., Faure, signals recognized by chemometric treatment of molecular strati- G., Rougerie, F., 1996. Deglacial sea level record from Tahiti corals graphic data. Organic Geochemistry 10, 649}660. and the timing of global meltwater discharge. Nature 382, 241}244. Brathauer, U., Abelmann, A., 1999. Late Quaternary variations in sea Bard, E., Hamelin, B., Fairbanks, R.G., Zindler, A., 1990a. Calibration surface temperatures and their relationship to orbital forcing re- of the C timescale over the past 130,000 years using mass spectro- corded in the Southern Ocean (Atlantic sector). Paleoceanography metric U-Th ages from Barbados corals. Nature 345, 405}410. 14, 135}148. Bard, E., Hamelin, B., Fairbanks, R.G., 1990b. U/Th ages obtained by Broccoli, A.J., 2000. Tropical cooling at the Last Glacial Maximum: an mass spectrometry in corals from Barbados: sea level during the atmosphere-mixed layer ocean model simulation. Journal of Cli- past 130,000 years. Nature 346, 456}458. mate 13, 951. Bard, E., Rostek, F., Sonzogni, C., 1997. Inter-hemispheric synchrony of Broccoli, A., Marciniak, E.P., 1996. Comparing simulated glacial cli- the last deglaciation inferred from alkenone palaeothermometry. mate and paleodata: a reexamination. Paleoceanography 11, 3}14. Nature 385, 707}710. Broecker, W.S., 1986. Oxygen isotope constraints on surface ocean Bard, E., Rostek, F., Turon, J.-L., Gendreau, S., 2000. Hydrological temperatures. Quaternary Research 26, 121}134. impact of Heinrich events in the subtropical northeast Atlantic. Broecker, W.S., Oppo, D., Peng, T.H., Curry, W., Andree, M., WoK l#i, Science 289, 1321}1324. W., Bonani, G., 1988. Radiocarbon-based chronology for the Barnes, D.J., Taylor, R.B., Lough, J.M., 1995. On the inclusion of trace O/O record for the last deglaciation. Paleoceanography 3, materials into massive coral skeleton. Part II: distortions in skeletal 509}515. records of annual climate cycles due to growth processes. Journal of Broecker, W.S., Andree, M., Bonani, G., Woel#i, W., Oeschger, H., Experimental Marine Biology and Ecology 194, 251}275. Klas, M., Mix, A.C., Curry, W., 1998. Preliminary estimates for the Barrows, T.T., Juggins, S., De Deckker, P., Thiede, J., Martinez, J.I., radiocarbon age of deep water in the glacial ocean. Paleoceanogra- 2000. Sea-surface temperatures of the southwest Paci"c Ocean phy 3, 659}669. during the Last Glacial Maximum. Paleoceanography 15, 95}109. Brown, S.J., Elder"eld, H., 1996. Variations in Mg/Ca and Sr/Ca ratios BeH , A.W.H., Duplessy, J.-C., 1976. Subtropical convergence #uctuations of planktonic foraminifera caused by depositional dissolution: evid- and Quaternary climates in the middle latitudes of the Indian ence of shallow Mg-dependent dissolution. Paleoceanography 11, Ocean. Science 194, 419}422. 543}551. Beaufort, L., Bassinot, F., Vincent, E., 1999. Primary production re- Burckle, L.H., 1984. Diatom distribution and paleoceanographic recon- sponse to orbitally induced variations of the southern oscillation in struction in the Southern Ocean: present and Last Glacial Max- the equatorial Indian Ocean. In: Abrantes, F., Mix, A.C. (Eds.), imum. Marine Micropaleontology 9, 241}261. Reconstructing Ocean History: a Window into the Future. Plenum, Burckle, L.H., Mortlock, R., 1998. Sea-ice extent in the Southern Ocean New York, pp. 245}271. during the Last Glacial Maximum: another approach to the prob- Beck, J.W., ReH cy, J., Taylor, F., Edwards, L., Cabioch, G., 1997. Abrupt lem. Annals of Glaciology 27, 302}304. changes in early Holocene tropical sea surface temperatures derived Burckle, L.H., Robinson, D., Cooke, D., 1982. Reappraisal of sea-ice from coral records. Nature 385, 705}707. distribution in the Atlantic and Paci"c sectors of the Southern Bender, M.L., Malize, B., Orchardo, J., Sowers, T., Jouzel, J., 1999. High Ocean at 18,000 yr BP. Nature 299, 435}437. precision correlations of Greenland and Antarctic ice core records Bush, A.B.G., Philander, S.G.H., 1998. The role of ocean-atmosphere over the last 100 kyr. In: Clark, P., Webb, R., Keigwin, L. (Eds.), interactions in tropical cooling during the last glacial maximum. Mechanisms of Global Climate Change at Millennial Time Scales, Science 279, 1341}1344. Geophysical Monograph, Vol. 112. American Geophysical Union, Cardinal, D., Hamelin, B., Bard, E., PaK tzold, J., 2000. Sr/Ca, U/Ca and Washington, DC, pp. 149}164. dO records in massive corals from Bermuda: relationships with Benthien, A., MuK ller, P.J., 2000. Anomalously low alkenone temper- sea-surface temperatures. Chemical Geology, in press. atures caused by lateral particle and sediment transport in the Cayre, O., Bard, E., 1999. The last deglaciation in the Eastern Arabian Malvinas Current region, western Argentine Basin. Deep-Sea Re- Sea: planktonic foraminiferal and alkenone records. Quaternary search I, in press. Research 52, 337}342. Berger, W.H., Gardner, J.V., 1975. On the determination of Pleistocene Cayre, O., Beaufort, L., Vincent, E., 1999. Paleoproductivity in the temperatures from planktonic foraminifera. Journal of Equatorial Indian Ocean for the last 260,000 yr: a transfer function Foraminiferal Research 5, 102}113. based on planktonic foraminifera. Quaternary Science Reviews 18, Betts, A.K., Ridgeway, W., 1992. Tropical boundary layer equilibrium 839}857. in the last ice age. Journal of Geophysical Research 97, Chapman, M.R., Shackleton, N.J., Zhao, M., Eglinton, G., 1996. Faunal 2529}2534. and alkenone reconstructions of subtropical North Atlantic surface Blunier, T., Chappellaz, J., Schwander, J., DaK llenbach, A., Stau!er, B., hydrography and paleotemperature over the last 28 kyr. Stocker, T., Raynaud, D., Jouzel, J., Clausen, H.B., Hammer, C.U., Paleoceanography 11, 343}357. Johnsen, S.J., 1998. Asynchrony of Antarctic and Greenland climate Charles, C.D., Lynch-Stieglitz, J., Ninnemann, U.S., Fairbanks, R.G., change during the last glacial period. Nature 394, 739}743. 1996. Climate connections between the hemispheres revealed by Blunier, T., Schwander, J., Stau!er, B., Stocker, T., DaK llenbach, A., deep-sea sediment core/ice core correlations. Earth and Planetary IndermuK hle, A., Tschumi, J., Chappellaz, J., Raynaud, D., Barnola, Science Letters 142, 19}27. J.M., 1997. Timing of the Antarctic Cold Reversal and the atmo- Clark, P.U., Alley, R.B., Keigwin, L.D., Licciardi, J.M., Johnsen, S.J., spheric CO increase with respect to the Younger Dryas event. Wang, H., 1996. Origin of the "rst global meltwater pulse following Geophysical Research Letters 24, 2683}2686. the last glacial maximum. Paleoceanography 11, 563}577. Boiseau, M., Cornu, H., Turpin, L., Juillet-Leclerc, A., 1997. Sr/Ca and Clark, P.U., Alley, R.B., Pollard, D., 1999. Northern hemisphere ice- dO measured from Acropora nobilis and Porites lutea: is Sr/Ca sheet in#uences on global climate change. Science 286, 1104}1111. paleothermometry always reliable?. Comptes Rendus de l'AcadeH mie Clark, P.U., Mix, A.C., 2000. Ice sheets by volume. Nature 406, des Sciences 325, 747}752. 689}690. 652 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

CLIMAP, 1976.The surface of the Ice-Age Earth. Science 191, deVilliers, S., Nelson, B.K., Chivas, A.R., 1995. Biological controls on 1131}1137. coral Sr/Ca and dO reconstructions of sea surface temperatures. CLIMAP, 1981. Seasonal reconstruction of the Earth's surface at the Science 269, 1247}1249. last glacial maximum. Geological Society of America, Map and deVilliers, S., Shen, G.T., Nelson, B.K., 1994. The Sr/Ca-temperature Chart Series, Vol. C36. relationship in coralline aragonite: in#uence of variability in (Sr/Ca) Cline, R.M., Hays, J.D. (Eds.), 1976. Investigation of Late Quaternary seawater and skeletal growth parameters. Geochimica et Cos- Paleoceanography and Paleoclimatology, Memoir 145, Geological mochimica Acta 58, 197}208. Society of America, Boulder, CO, 464 pp. Dole, M., Jenks, G., 1944. Isotopic composition of photosynthetic COHMAP, 1988. Climatic changes of the last 18,000 years: observa- oxygen. Science 100, 409. tions and model simulations. Science 241, 1043}1052. Doose, H., Prahl, F.G., Lyle, M.W., 1997. Biomarker temperature Colinvaux, P.A., De Oliveira, P.E., Bush, M.B., 2000. Amazonian and estimates for modern and last glacial surface waters of the California neotropical plant communities on glacial time-scales: the failure of Current system between 333 and 423N. Paleoceanography 12, 615}622. the aridity and refuge hypotheses. Quaternary Science Reviews 19, Duplessy, J.C., Bard, E., Arnold, M., Shackleton, N.J., Duprat, J., 141}169. Labeyrie, L.D., 1991. How fast did the ocean-atmosphere system Conte, M.H., Thompson, A., Lesley, D., Harris, R.P., 1998. Genetic and run during the last deglaciation? Earth and Planetary Science physiological in#uences on the alkenone/alkenoate versus growth Letters 103, 27}40. temperature relationship in Emiliania huxleyi and Gephyrocapsa Dwyer, G.S., Cronin, T.M., Baker, P.A., Raymo, M.E., Buzas, J.S., oceanica. Geochimica et Cosmochimica Acta 62, 51}68. Corre`ge, T., 1993. North Atlantic deepwater temperature change Cooke, D.W., Hays, J.D., 1982. Estimates of Antarctic Ocean seasonal during Late Pliocene and Late Quaternary climatic cycles. Science sea-ice cover during glacial intervals. In: Craddock, C. (Ed.), Antarc- 270, 1347}1351. tic Geoscience. Univ. Wisconsin Press, Madison, WI, pp. Edwards, R.L., Beck, J.W., Burr, G.S., Donahue, D.J., Chappell, J.M.A., 1017}1025. Bloom, A.L., Dru!el, E.R.M., Taylor, F.W., 1993. A large drop in Cortijo, E., Labeyrie, L., Vidal, L., Vautravers, M., Chapman, M., atmospheric C/C and reduced melting in the Younger Dryas, Dupessy, J.C., Elliot, M., Arnold, M., Turon, J.L., Au!ret, G., 1997. documented with Th ages of corals. Science 260, 962}968. Changes in sea surface hydrology associated with Heinrich event Elder"eld, H., Ganssen, G., 2000. Past temperature and dO of surface 4 in the North Atlantic Ocean between 403 and 603N. Earth and seawaters inferred from foraminiferal Mg/Ca ratios. Nature 405, Planetary Science Letters 146, 29}45. 442}445. Cross, T.S., Cross, B.W., 1983. U, Sr and Mg in Holocene and Pleis- Elliot, M., Labeyrie, L., Bond, G., Cortijo, E., Turon, J.L., Tisnerat, N., tocene corals. Journal of Sedimentary Petrology 53, 587}594. Duplessy, J.-C., 1998. Millennial-scale iceberg discharges in the Crosta, X., Pichon, J.-J., Burckle, L.H., 1998a. Application of modern Irminger Basin during the last glacial period: relationship with the analog technique to marine Antarctic diatoms: reconstruction of Heinrich events and environmental settings. Paleoceanography 13, maximum sea ice extent at the Last Glacial Maximum. 433}446. Paleoceanography 13, 286}297. Emiliani, C., 1955. Pleistocene temperatures. Journal of Geology 65, Crosta, X., Pichon, J.-J., Burckle, L.H., 1998b. Reappraisal of Antarctic 538}576. seasonal sea ice at the last Glacial Maximum. Geophysical Research Epstein, B.L., D'Hondt, S., Quinn, J.C., Zhang, J., Hargraves, P.E., Letters 25, 2703}2706. 1998. An e!ect of dissolved nutrient concentration on alkenone Crowley, T.J., 2000. CLIMAP SSTs re-revisited. Climate Dynamics 16, based temperature estimates. Paleoceanography 12, 122}126. 241}255. Fairbanks, R.G., 1989. A 17,000-year glacio-eustatic sea level record: Crowley, T.J., Baum, S.K., 1997. E!ect of vegetation on an ice-age in#uence of glacial melting rates on the Younger Dryas event and climate model simulation. Journal of Geophysical Research 102, deep ocean circulation. Nature 342, 637}647. 16,463}16,480. Farrera, I., Harrison, S.P., Prentice, I.C., Ramstein, G., Guiot, J., Bar- Cu!ey, K.M., Clow, G.D., Alley, R.B., Stuiver, M., Waddington, E.D., tlein, P.J., Bonne"lle, R., Bush, M., Cramer, W., vonGrafenstein, U., Saltus, R.W., 1995. Large Arctic temperature change at the Wiscon- Holmgren, K., Hooghiemstra, H., Hope, G., Jolly, D., Lauritzen, sin-Holocene glacial transition. Science 270, 455}458. S.-E., Ono, Y., Pinot, S., Stute, M., Yu, G., 1999. Tropical climates at Dansgaard, W., Johnsen, S.J., Clausen, H.B., Dahl-Jensen, D., Gundes- the Last Glacial Maximum: a new synthesis of terrestrial palaeocli- trup, N.S., Hammer, C.U., Hvidberg, C.S., Ste!ensen, J.P., Svein- mate data. I. Vegetation, lake-levels, and geochemistry. Climate bjornsdottir, A.E., Jouzel, J., Bond, G.C., 1993. Evidence for general Dynamics 15, 823}856. instability of past climate from a 250-kyr ice-core record. Nature Fischer, H., Wahlen, M., Smith, J., Mastroianni, D., Deck, B., 1999. Ice 264, 218}220. core records of atmospheric CO around the last three glacial Delaygue, G., Jouzel, J., Duplessy, J.-C., 2000. Oxygen 18-salinity rela- terminations. Science 283, 1712}1714. tionship simulated by an oceanic general circulation model. Earth Fleming, K., Johnston, P., Zwartz, D., Yokoyama, Y., Lambeck, K., and Planetary Science Letters 178, 113}123. Chappell, J., 1998. Re"ning the eustatic sea-level curve since the deVernal, A., Hillaire-Marcel, C., 2000. Sea-ice cover, sea-surface salin- Last Glacial Maximum using far- and intermediate-"eld sites. Earth ity, and halo-/thermocline structure of the northwest North Atlan- and Planetary Sciences Letters 163, 327}342. tic: modern versus full glacial conditions. Quaternary Science Ganopolski, A., Rahmstorf, S., Petoukhov, V., Claussen, M., 1998. Reviews 19, 65}85. Simulation of modern and glacial climates with a coupled global deVernal, A., Pedersen, T.F., 1997. Micropaleontology and palynology model of intermediate complexity. Nature 391, 351}356. of core PAR87A-10: a 23,000-year record of paleoenvironmental Gates, L., 1976. Modeling the ice-age climate. Science 191, 1138}1144. changes in the Gulf of Alaska, northeast North Paci"c. Geitzenauer, K.R., Roche, M.B., McIntyre, A., 1976. Modern Paci"c Paleoceanography 12, 821}830. coccolith assemblages: derivation and application to late Pleis- deVernal, A., Turon, J.-L., Guiot, J., 1994. Dino#agellate cyst distribu- tocene paleotemperature analysis. Geological Society of America tion in high-latitude marine environments and quantitative recon- Memoir 145, 423}448. struction of sea-surface salinity, temperature, and seasonality. Greegor, R.B., Pingitore, N.E., Lytle, F.W., 1997. Strontianite in coral Canadian Journal of Earth Science 31, 48}62. skeletal aragonite. Science 275, 1452}1454. deVilliers, S., 1999. Seawater strontium and Sr/Ca variability in the Grootes, P.M., Stuiver, M., White, J.W.C., Johnsen, S.J., Jouzel, J., 1993. Atlantic and Paci"c oceans. Earth and Planetary Science Letters Comparison of oxygen isotope records from the GISP2 and GRIP 171, 623}634. Greenland ice cores. Nature 366, 552}554. A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 653

Guilderson, T.P., Fairbanks, R.G., Rubenstone, J.L., 1994. Tropical Ip, Y.K., Lim, A.L.L., 1991. Are calcium and strontium transported by temperature variations since 20,000 years ago: modulating inter- the same mechanism in the hermatypic coral Galaxea fascicularis?. hemispheric climate change. Science 263, 663}665. Journal of Experimental Biology 159, 507}513. Guiot, J., 1990. Methods and programs of statistics for paleoclimatol- Johnsen, S.J., Clausen, H.B., Dansgaard, W., Gundestrup, N.S., Ham- ogy and paleoecology. Quanti"cation des Changements Climat- mer, C.U., Andersen, U., Andersen, K.K., Hvidberg, C.S., Dahl- iques: meH thodes et Programmes, Monograph 1, CNRS-France, Jensen, D., Ste!ensen, J.P., Shoji, H., Sveinbjornsdottir, A.E., White, 253 pp. J.W.C., Jouzel, J., Fisher. D., 1997. The dO record along the Hammer, C.U., Andersen, K.K., Clausen, H.B., Dahl-Jensen, D., Hvid- Greenland Ice Core Project deep ice core and the problem of berg, C.S., Iversen, P., 1997. The stratigraphic dating of the GRIP possible Eemian climatic instability. Journal of Geophysical Re- ice core. Special Report of the Geophysical Department, Niels Bohr search 102, 26,397}26,410. Institute for Astronomy, Physics and Geophysics, University of Johnsen, S.J., Dahl-Jensen, D., Dansgaard, W., Gundestrup, N., 1995. Copenhagen, GRIP-GISP CD-ROM ("le gripstrt). Greenland paleotemperatures derived from GRIP bore hole tem- Hammer, C.U., Clausen, H.B., Langway, C.C., 1994. Electrical conduct- perature and ice core isotope pro"les. Tellus 47B, 624}629. ivity method (ECM) and stratigraphic dating of the Byrd station ice Johnsen, S.J., Dansgaard, W., Claussen, H.B., Langway, C.C., 1972. core Antarctica. Annals of Glaciology 20, 115}120. Oxygen isotope pro"les through the Antarctic and Greenland ice Hanebuth, T., Stattegger, K., Grootes, P.M., 2000. Rapid #ooding of the sheets. Nature 235, 429}434. Sunda Shelf: a late-glacial sea-level record. Science 288, 1033}1035. Keigwin, L.D., 1998. Glacial-age hydrography of the far northwest Hart, S.R., Cohen, A.L., 1996. An ion probe study of annual cycles of Paci"c Ocean. Paleoceanography 13, 323}339. Sr/Ca and other trace elements in corals. Geochimica et Cos- Kennett, J.P., Shackleton, N.J., 1975. Laurentide ice sheet meltwater mochimica Acta 60, 3075}3084. recorded in Gulf of Mexico deep-sea cores. Science 188, 147}150. Hastings, D.W., Russell, A.D., Emerson, S.R., 1998. Foraminiferal mag- Kipp, N.G., 1976. New transfer function for estimating past sea surface nesium in Globigerinoides sacculifer as a paleotemperature proxy. conditions from seabed distribution of planktonic forminiferal as- Paleoceanography 13, 161}169. semblages in the North Atlantic. In: Cline, R.M., Hays, J.D. (Eds.), Hays, J.D., Lozano, J.A., Shackleton, N.J., Irving, G., 1976. Reconstruc- Investigation of Late Quaternary Paleoceanography and Paleo- tion of the Atlantic and western Indian Ocean sectors of the 18,000 climatology, Mem. 145. Geological Society of America, Boulder, BP Antarctic Ocean. Geological Society of America Memoir 145, CO, pp. 3}41. 337}372. Kitigawa, H., van der Plicht, J., 1998. A 40,000-year varve chronology Heinrich, H., 1988. Origin and consequences of cyclic ice rafting in the from Lake Suigetsu, Japan: extension of the C calibration curve. northeast Atlantic Ocean during the past 130,000 years. Quaternary Radiocarbon 40, 505}515. Research 29, 143}152. Kohfeld, K.E., Fairbanks, R.G., Smith, S.L., Walsh, I.D, 1996. Neo- Hendy, I.L., Kennett, J.P., 1999. Latest Quaternary North Paci"c globoquadrina pachyderma (sinistral coiling) as paleoceanographic surface-water responses imply atmosphere-driven climate instabil- tracers in polar oceans: evidence from Northeast Water Polynya ity. Geology 27, 291}294. plankton tows, sediment traps, and surface sediments. Paleoceano- Herbert, T., Schu!ert, J.D., Thomas, D., Lange, C., Weinheimer, A., graphy 11, 679}700. Peleo-Alampay, A., Herguera, J.-C., 1998. Depth and seasonality of Kutzbach, J.E., Guetter, P.J., 1986. The in#uence of changing orbital alkenone production along the California margin inferred from parameters and surface boundary conditions on climate simulations a core top transect. Paleoceanography 13, 263}271. for the past 18,000 years. Journal of Atmospheric Science 43, Herbert, T.D., Yasuda, M., Burnett, C., 1995. Glacial}interglacial sea- 1726}1759. surface temperature record inferred from alkenone indices, Site 893, Labeyrie, L.D., Duplessy, J.-C., Blanc, P.L., 1987. Variations in mode of Santa Barbara Basin. In: Kennett, J.P., Baldauf, J.G., Lyle, M. formation and temperature of oceanic deep waters over the past (Eds.), Proceedings of the Ocean Drilling Program, Vol. 146 (pt. 2), 125,000 years. Nature 327, 477}482. College Station, TX, pp. 257}264. Lambeck, K., Yokoyama, Y., Johnson, P., Purcell, A., 2000. Global ice Hostetler, S.W., Mix, A.C., 1999. Reassessment of ice-age cooling of the volumes at the Last Glacial Maximum and early Lateglacial, 2000. tropical ocean and atmosphere. Nature 399, 673}676. Earth and Planetary Science Letters 181, 513}527. Hughes, T.J., Denton, G.H., Andersen, B.G., Schilling, D.H., Fastook, Lea, D.W., Mashiotta, T.A., Spero, H.J., 1999. Controls on magnesium J.L., Lingle, C.S, 1981. The last great ice sheets: a global view. In: and strontium uptake in planktonic foraminifera determined by live Denton, G.H., Hughes, T.J. (Eds.), The Last Great Ice Sheets. Wiley, culturing. Geochimica et Cosmochimica Acta 63, 2369}2379. New York, pp. 275}317. Lea, D.W., Pak, D.K., Spero, H.J., 2000. Climate Impact of late Quater- Hutson, W.H., 1977. Transfer functions under no-analog conditions: nary equatorial Paci"c sea-surface temperature variations. Science experiments with Indian Ocean planktonic foraminifera. Quater- 289, 1719}1724. nary Research 8, 355}367. Lee, K.E., Slowey, N.C., 1999. Cool surface waters of the subtropical Hutson, W.H., 1980. The Agulhas Current during the late Pleistocene: North Paci"c Ocean during the last glacial. Nature 318, 556}557. analysis of modern faunal analogs. Science 207, 64}66. Lozano, J.A., Hays, J.D., 1976. Relationship of Radiolarian assemblages Imbrie, J., Kipp, N.G., 1971. A new micropaleontological method for to sediment types and physical oceanography in the Atlantic and quantitative paleoclimatology: application to a late Pleistocene western Indian Ocean sectors of the Antarctic Ocean. In: Cline, Caribbean core. In: Turekian, K.K. (Ed.), Late Cenozoic Glacial R.M., Hays, J.D. (Eds.), Investigation of Late Quaternary Ages. Yale Univ. Press, New Haven, CN, pp. 71}182. Paleoceanography and Paleoclimatology, Mem. 145. Geological Imbrie, J., McIntyre, A., Mix, A.C., 1989. Oceanic response to orbital Soceity of America, Boulder, CO, pp. 303}336. forcing in the late Quaternary: observational and experimental Lund, D.C., Mix, A.C., 1998. Millennial-scale deep water oscillations: strategies. In: Berger, A., Schneider, S., Duplessy, J.-C. (Eds.), re#ections of the North Atlantic in the deep Paci"c from 10 to 60 ka. Climate and Geosciences. Kluwer Academic, Dordrecht, Paleoceanography 13, 10}19. pp. 121}164. Madureira, L.A.S., Conte, M.H., Eglinton, G., 1995. Early diagenesis of Imbrie, J., et al., 1993. On the structure and origin of major glaciation lipid biomarker compounds in North Atlantic sediments. cycles 2. The 100,000-year cycle. Paleoceanography 8, 699}735. Paleoceanography 10, 627}642. Ip, Y.K., Krishnaveni, P., 1991. Incorporation of Strontium (Sr>) Malmgren, B.A., Nordlund, U., 1997. Application of arti"cial neural into the skeleton of the hermatypic coral Galaxea fascicularis. The networks to paleoceanographic data. Palaeogeography Palaeo- Journal of Experimental Zoology 258, 273}276. climatology and Palaeoecology 136, 359}373. 654 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

Manabe, S., Hahn, D.G., 1977. Simulation of the tropical climate of an Mix, A.C., Ruddiman, W.F., McIntyre, A., 1986b. Late Quaternary ice age. Journal of Geophysical Research 82, 3889}3911. paleoceanography of the tropical Atlantic, II : the seasonal cycle of Marlowe, I.T., Green, J.C., Neal, A.C., Brassell, S.C., Eglinton, G., sea-surface temperatures, 0-20,000 years B.P. Paleoceanography 1, Course, P.A., 1984. Long chain (n-C37-C39) alkenones in the Prym- 339}353. nesiophyceae: Distributions of alkenones and other lipids and their Mix, A.C., Morey, A.E., Pisias, N.G., Hostetler, S., 1999b. Foraminiferal taxonomic signi"cance. British Phycology Journal 19, 203}216. faunal estimates of paleotemperatures: circumventing the no-analog Martin, P.A., Lea, D.W., Mashiotta, T.A., Papenfuss, T., Sarnthein, M., problem yields cool ice-age tropics. Paleoceanography 14, 350}359. 2000. Variation of foraminiferal Sr/Ca over Quaternary glacial- Mix, A.C., Lund, D.C., Pisias, N.G., BodeH n, P., Bornmalm, L., Lyle, M., interglacial cycles: evidence for changes in mean ocean Sr/Ca? Pike, J., 1999a. Rapid climate oscillations in the Northeast Paci"c Geochemistry Geophysics and Geosystems 1, 6. during the last deglaciation re#ect northern and southern hemi- Mashiotta, T.A., Lea, D.W., Spero, H.J., 1999. Glacial-interglacial cha- sphere sources. In: Clark, P., Webb, R., Keigwin, L. (Eds.), Mecha- nges in Subantarctic sea surface temperature and dO -water using nisms of Global Climate Change at Millennial Time Scales, foraminiferal Mg. Earth and Planetary Science Letters 170, Geophysical Monograph, Vol. 112. American Geophysical Union, 417}432. Washington, DC, pp. 127}148. McIntyre, A., 1967. Coccoliths as paleoclimatic indicators of Pleis- Mol"no, B., Kipp, N.G., Morley, J.J., 1982. Comparison of tocene glaciations. Science 158, 1314}1317. foraminiferal, coccolithophorid, and radiolarian paleotemperature McIntyre, A., Kipp, N.G., BeH , A.W.H., Crowley, T., Kellogg, T., Gar- equations: assemblage coherency and estimate concordancy. Quat- dner, J.V., Prell, W., Ruddiman, W.F., 1976. Glacial North Atlantic ernary Research 17, 279}313. years ago: a CLIMAP reconstruction. In: Cline, R.M., Hays, J.D. Moore Jr., T.C., Burckle, L.H., Geitzenauer, K., Luz, B., Molina-Cruz, (Eds.), Investigation of Late Quaternary Paleoceanography and A., Robertson, J.H., Sachs, H., Sancetta, C., Thiede, J., Thompson, Paleoclimatology, Mem. 145. Geol. Soc. Am, Boulder, CO, pp. P., Wenkam, C., 1980. The reconstruction of sea surface temper- 43}76. atures in the Paci"c Ocean of 18,000 B.P. Marine Micropaleontol- Meese, D.A., Gow, A.J., Alley, R.B., Zielinski, G.A., Grootes, P.M., ogy 5, 215}247. Ram, M., Taylor, K.C., Mayewski, P.A., Bolzan, J.F., 1997. The Moore Jr., T.C., Hutson, W.H., Kipp, N.G., Hays, J.D., Prell, W.L., Greenland Ice Sheet Project 2 depth-age scale: methods and results. Thompson, P., Boden, G., 1981. The biological record of the ice age Journal of Geophysical Research 102, 26,411}26,423. ocean. Palaeogeography Palaeolimatology and Palaeoecology 35, Meese, D.A., Alley, R.B., Fiacco, R.J., Germani, M.S., Gow, A.J., 357}370. Grootes, P.M., Illing, M., Mayewski, P.A., Morrison, M.C., Ram, Mulitza, S., Niebler, H.S., DuK rkoop, A., Hale, W., Wefer, G., 1997. M., Taylor, K.C., Yang, Q., Zielinski, G.A., 1994. Preliminary Planktonic foraminifera as recorders of past surface-water strati"- depth-age scale of the GISP2 ice core. Special CRREL Report 94-1, cation. Geology 25, 335}338. US. MuK ller, P.J., Kirst, G., Ruhland, G., von Storch, I., Rosell-MeleH , A., Miao, Q., Thunell, R.C., Anderson, D.M., 1994. Glacial-Holocene car- 1998. Calibration of the alkenone paleotemperature index Y bonate dissolution and sea surface temperatures in the South China U based on core-tops from the eastern South Atlantic and global and Sulu Seas. Paleoceanography 9, 269}290. ocean (603N-603S). Geochimica et Cosmochimica Acta 62, Min, G.R., Edwards, R.L., Taylor, F.W., Recy, J., Gallup, C.D., Beck, 1757}1772. J.W., 1995. Annual cycles of U/Ca in coral skeletons and U/Ca Mulvaney, R., RoK thlisberger, R., Wol!, E.W., Sommer, S., Schwander, thermometry. Geochimica et Cosmochimica Acta 59, 2025}2042. J., Hutterli, M.A., Jouzel, J., 2000. The transition from the last Mix, A.C., 1987. The oxygen-isotope record of glaciation. In: Ruddi- glacial period in inland and near-coastal Antarctica. Geophysical man, W.F, Wright, H.E. (Eds.), North America and Adjacent Research Letters 27, 2673}2676. Oceans during the Last Deglaciation The Geology of North Amer- National Oceanographic Data Center, 1999. World Ocean Atlas 1998 ica, Vol. K-3. Geological Society of America, Boulder, CO, pp. (WOA98), CD-ROM, Ocean Climate Laboratory, NODC, 111}135. www.nodc.noaa.gov. Mix, A.C., 1989. Pleistocene paleoproductivity: evidence from organic NuK rnberg, D., Bijma, J., Hemleben, C., 1996. Assessing the reliability of carbon and foraminiferal species. In: Berger, W.H., Smetacek, V.S., magnesium in foraminiferal calcite as a proxy for water mass tem- Wefer, G. (Eds.), Productivity of the Ocean, Past and Present. perature. Geochimica et Cosmochimica Acta 60, 803}814. Wiley, New York, pp. 313}340. NuK rnberg, D., MuK ller, A., Schneider, R.R., 2000. Paleo-sea surface Mix, A.C., 1992. The marine oxygen-isotope record: constraints on temperature calculations in the equatorial east Atlantic from timing and extent of ice growth events (120-65 ka). In: Clark, P.U., Mg/Ca ratios in planktic foraminifera: a comparison to sea surface Y Lea, P.D. (Eds.), The Last Interglacial}Glacial Transition in North temperature estimates from U, oxygen isotopes, and foraminiferal America. Special Paper 270. Geological Society of America, Boul- transfer function. Paleoceanography 15, 124}134. der, CO, pp. 19}30. Okhouchi, N., Kawamura, K., Nakamura, T., Taira, A., 1994. Small Mix, A.C., Morey, A.E., 1996. Climate Feedback and Pleistocene vari- changes in the sea surface temperature during the last 20,000 years: ations in the Atlantic South Equatorial Currents. In: Wefer, G., molecular evidence from the western tropical Paci"c. Geophysics Berger, W.H., Siedler, G., Webb, D. (Eds.), The South Atlantic: Research Letters 20, 2207}2210. Present and Past Circulation. Springer-Verlag, Berlin, pp. 503}525. Okhouchi, N., Kawamura, K., Kawahata, H., Okada, H., 1999. Depth Mix, A.C., Pisias, N.G., 1988. Oxygen isotope analyses and deep-sea ranges of alkenone production in the central Paci"c Ocean. Global temperature changes: implications for rates of oceanic mixing. Na- Biogeochemical Cycles 13, 695}704. ture 331, 249}251. Ortiz, J.D., Mix, A.C., 1997. Comparison of Imbrie-Kipp transfer Mix, A.C., Ruddiman, W.F., 1984. Oxygen-isotope analyses and Pleis- function and modern analog temperature estimates using sediment tocene ice volumes. Quaternary Research 21, 1}20. trap and core top foraminiferal faunas. Paleoceanography 12, Mix, A.C., Ruddiman, W.F., 1985. Structure and timing of the last 175}190. deglaciation: oxygen-isotope evidence. Quaternary Science Reviews Ortiz, J.D., Mix, A.C., Hostetler, S., Kashgarian, M., 1997. The Califor- 4, 59}108. nia Current of the last glacial maximum: reconstruction at 423N Mix, A.C., Ruddiman, W.F., McIntyre, A., 1986a. Late Quaternary based on multiple proxies. Paleoceanography 12, 191}206. paleoceanography of the tropical Atlantic, 1, Spatial variability of Parker, F., Berger, W.H., 1971. Faunal and solution patterns of plank- annual mean sea surface temperatures, 0-20,000 years B.P. tonic foraminifera in surface sediments of the South Paci"c. Deep- Paleoceanography 1, 43}66. Sea Research 18, 73}107. A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 655

Pastouret, L., Chamley, H., Delibrias, G., Duplessy, J.C., Thiede, J., Prahl, F.G., Wakeham, S.G., 1987. Calibration of unsaturation patterns 1978. Late Quaternary climatic changes in western tropical Africa in long-chain ketone compositions for paleotemperature assess- deduced from deep-sea sedimentation o! the Niger delta. Oceanol- ment. Nature 330, 367}369. ogy Acta. 1, 217}232. Prell, W.L., 1985. The stability of Low-Latitude Sea-Surface Temper- Peltier, W.R., 1994. Ice age paleotopography. Science 265, 195}201. atures: an Evaluation of the CLIMAP Reconstruction with Empha- Peltier, W.R., 1998a. `Implicit icea in the global theory of glacial sis on the Positive SST Anomalies. Technical Report 25, United isostatic adjustment. Geophysical Research Letters 25, 3955}3958. States Dept. of Energy, Washington, DC, pp. 1}60. Peltier, W.R., 1998b. Postglacial variations in the level of the sea: Rahmstorf, S., 1995. Bifurcations of the Atlantic thermohaline circula- implications for climate dynamics and solid-earth geophysics. Re- tion in response to changes in the hydrological cycle. Nature 378, views of Geophysics 36, 603}689. 145}149. Peng, T.-H., Broecker, W.S., Kipphut, G., Shackleton, N., 1977. Benthic Ramstein, G., Joussaume, S., 1995. Sensitivity experiments to sea mixing in deep sea cores as determined by C dating and its surface temperatures, sea-ice extent, and ice-sheet reconstruc- implications regarding climate stratigraphy and the fate of fossil fuel tions for the last glacial maximum. Annals of Glaciology 21, CO. In: Andersen, N., Malaho!, A. (Eds.), The Fate of Fossil Fuel 343}347. CO in the Ocean. Plenum, New York, pp. 355}374. Ravelo, A.C., Andreasen, D.H., 1999. Using planktonic foraminifera as Petit and 18 others, 1999. Climate and atmospheric history of the past monitors of the tropical surface ocean. In: Abrantes, F., Mix, A.C. 420,000 years from the Vostok ice core, Antarctica. Nature 399, (Eds.), Reconstructing Ocean History: a Window into the Future. 429}436. Plenum, New York, pp. 217}243. Peteet, D., Del Genio, A., Lo, K.-W., 1997. Sensitivity of northern Ravelo, A.C., Fairbanks, R.G., Philander, S.G.H., 1990. Reconstructing hemisphere air temperatures and snow expansion to North Paci"c tropical Atlantic hydrography using planktonic foraminifera and an sea surface temperatures in the Goddard Institute for Space Studies ocean model. Paleoceanography 5, 409}431. general circulation model. Journal of Geophysics Research 102, Rind, D., Peteet, D., 1985. Terrestrial conditions at the Last Glacial 23,781}23,791. Maximum and CLIMAP sea-surface temperature estimates: are Peterson, G.M., Webb, T.III, Kutzbach, J.E., van der Hammen, T., they consistent? Quaternary Research 24, 1}22. Wijmstra, T., Street, F.A., 1979. The continental record of environ- Rohling, E.J., Bigg, G.R., 1998. Paleosalinity and dO: a critical assess- mental conditions at 18,00 yr B.P.: an initial evaluation. Quaternary ment. Journal of Geophysics Research 103, 1307}1318. Research 12, 47}82. Rosell-MeleH , A., 1998. Interhemispheric appraisal of the value of al- P#aumann, U., Duprat, J., Pujol, C., Labracherie, L., 1996. SIMMAX: kenone indices as temperature and salinity proxies in high-latitude a modern analog technique to deduce Atlantic sea surface temper- locations. Paleoceanography 13, 694}703. atures from planktonic foraminifera in deep-sea sediments. Rosell-MeleH , A., Bard, E., Emeis, K.-C., Farrimond, P., Grimalt, J., Paleoceanography 11, 15}35. MuK ller, P.J., Schneider, R.R., 1998. Project takes a new look at past Pichon, J.-J., Labeyrie, L.D., Bareille, G., Labracherie, M., Duprat, J., sea surface temperatures. EOS 79, 393}394. Jouzel, J., 1992. Surface water temperature changes in the high Rosell-MeleH , A., Bard, E., Grimalt, J., Harrison, I., Bouloubassi, I., latitudes of the southern hemisphere during the last glacial-inter- Comes, P., Emeis, K.-C., Epstein, B., Fahl, K., Farrimond, P., glacial cycle. Paleoceanography 7, 289}313. Fluegge, A., Freeman, K., Goni, M., GuK ntner, U., Hartz, D., Pichon, J.-J., Sikes, E.L., Hiramatsu, C., Robertson, L., 1998. Compari- Hellebust, S., Herbert, T., Ikehara, M., Ishiwatari, R., Kawamura, Y son of U and diatom assemblage sea surface temperature esti- K., Kenig, F., de Leeuw, J., Lehman, S., MuK ller, P., Ohkouchi, N., mates with atlas derived data in Holocene sediments from the Pancost, R.D., Prahl, F., Quinn, J., Rontani, J.-F., Rostek, F., Southern West Indian Ocean. Journal of Marine Systems 17, Rullkotter, J., Sachs, J., Sanders, D., Sawada, K., Schneider, R., 541}554. Schulz-Bull, D., Sikes, E., Ternois, Y., Versteegh, G., Volkman, J., Pinot, S., Ramstein, G., Harrison, S.P., Prentice, I.C., Guiot, J., Stute, Wakeham, S., 2000. Outcome of the "rst world-wide laborat- Y M., Joussaume, S., 1999. Tropical paleoclimates at the Last Glacial ory intercomparison of the U index and the quanti"cation Maximum: comparision of Paleoclimate Modeling Intercompari- of absolute alkenone abundances in sediment samples, sion Project (PMIP) simulations and paleodata. Climate Dynamics submitted. 15, 857}874. Rosell-MeleH , A., Comes, P., 1999. Evidence for a warm Last Glacial Pirazzoli, P.A., 1991. World Atlas of Holocene Sea-Level Changes, Maximum in theNordic seas or an example of shortcomings in Y Y Oceanography Series, Vol. 58. Elsevier, Amsterdam, 300 pp. U and U to estimate low sea surface temperature?. Pisias, N.G., Mix, A.C., 1997. Spatial and temporal oceanographic Paleoceanography 14, 770}776. variability of the eastern equatorial Paci"c during the late Pleis- Rosenthal, Y., Boyle, E., 1993. Factors controlling the #uoride content tocene: evidence from Radiolaria microfossils. Paleoceanography of planktic foraminifera. Geochimica et Cosmochimica Acta 57, 12, 381}393. 335}346. Pisias, N.G., Roelofs, A., Weber, M., 1997. Radiolarian-based transfer Rosenthal, Y., Lohmann, G.P., Lohmann, K.C., Sherrell, R.M., 2000. functions for estimating mean surface ocean temperatures and sea- Incorporation and preservation of Mg in Globigerinoides sacculifer: sonal range. Paleoceanography 12, 365}379. implications for reconstructing the temperature and O/Oof Popp, B.N., Kenig, F., Wakeham, S.G., Laws, E.A., Bidigare, R.R., 1998. seawater. Paleoceanography 15, 135}145. Does growth rate a!ect ketone unsaturation and intracellular car- Sabin, A.L., Pisias, N.G., 1996. Sea surface temperature changes in the bon isotopic variability in Emiliania huxleyi? Paleoceanography 13, Northeastern Paci"c ocean during the past 20,000 years and their 35}41. relationship to climate change in Northwestern North America. Prahl, F.G., Collier, R.B., Dymond, J., Lyle, M., Sparrow, M.A., 1993. Quaternary Research 46, 48}61. A biomarker perspective on prymnesiophyte productivity in the Sarnthein, M., Jansen, E., Weinelt, M., Arnold, M., Duplessy, J.-C., northeast Paci"c Ocean. Deep-Sea Research 40, 2061}2076. Erlenkeusser, H., Flat+y, A., Johannessen, G., Johannessen, T., Jung, Prahl, F.G., de Lange, G.J., Lyle, M., Sparrow, M.A., 1989. Post- S., Koc, N., Labeyrie, L., Maslin, M., P#aumann, U., Shulz, H., depositional stability of long-chain alkenones under contrasting 1995. Variations in Atlantic surface ocean paleoceanography 503} redox conditions. Nature 341, 434}437. 803N: a time-slice record of the last 30,000 years. Paleoceanography Prahl, F.G., Pisias, N.G., Sparrow, M.A., Sabin, A., 1995. Assessment of 10, 1063}1094. sea-surface temperature at 423N in the California Current over the Schmidt, G., 1999. Error analysis of paleosalinity calculations. last 30,000 years. Paleoceanography 10, 763}773. Paleoceanography 14, 422}429. 656 A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657

Schneider, R.R., MuK ller, P.J., Ruhland, G., Meinecke, G., Schmidt, H., Stephens, B.B., Keeling, R.F., 2000. The in#uence of Antarctic sea ice on Wefer, G., 1996. Late Quaternary surface temperatures and produc- glacial-interglacial CO variations. Nature 404, 171}174. tivity in the east-equatorial South Atlantic; response to changes Stoll, H.M., Schrag, D.P., 1998. E!ects of Quaternary sea level cycles on in trade/monsoon wind forcing and surface water advection. In: strontium in seawater. Geochimica et Cosmochimica Acta 62, Wefer, G., Berger, W.H., Siedler, G., Webb, D. (Eds.), The South 1107}1118. Atlantic: present and Past Circulation. Springer-Verlag, Berlin, Stuiver, M., Grootes, P.M., 2000. GISP2 oxygen isotope ratios. Quater- pp. 527}551. nary Research 53, 277}284. Schneider, R.R., MuK ller, P.J., Acheson, R., 1999. Atlantic alkenone Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen, K.A., sea-surface temperature records: low versus mid latitudes and di!er- Kromer, B., McCormick, G., van der Plicht, J., Spurk, M., 1998. ences between hemispheres. In: Abrantes, F., Mix, A.C. (Eds.), Re- INTCAL98 radiocarbon age calibration, 24,000-0 cal BP. Radiocar- constructing Ocean History: a Window into the Future. bon 40, 1041}1083. Kluwer/Plenum Publ, New York, Dordrecht, pp. 33}56. Stute, M., Forster, M., Frischkorn, H., Serejo, A., Clark, J.F., Schlosser, Schrag, D.P., 1999. Rapid determination of high-precision Sr/Ca ratios P., Broecker, W.S., Bonani, G., 1995. Cooling of tropical Brazil (53C) in corals and other marine carbonates. Paleoceanography 14, during the Last Glacial Maximum. Science 269, 379}383. 97}102. Swart, P.K., 1981. The strontium, magnesium and sodium composition Schrag, D.P., DePaolo, D.J., 1993. Determination of dO of seawater of recent scleractinian coral skeletons as standards for palaeoen- in the deep ocean during the Last Glacial Maximum. Paleoceano- vironmental analysis. Palaeogeography, Palaeoclimatology, graphy 8, 1}6. Palaeocology 34, 115}136. Schrag, D.P., Hampt, G., Murray, D.W., 1996. Pore #uid constraints on Teece, M.A., Getli!, J.M., Parkes, R.J., Leftley, J.W., Maxwell, J.R., the temperature and oxygen isotopic composition of the glacial 1998. Microbial degradation of the prymnesiophyte Emiliania hux- ocean. Science 272, 1930}1932. leyi under oxic and anoxic conditions as a model for early diagen- Shackleton, N.J., 1967. Oxygen isotopic analyses and Pleistocene tem- esis: long chain alkadienes, alkenones and alkyl alkenoates. peratures re-assessed. Nature 215, 15}17. Organics Geochemistry 29, 863}880. Shackleton, N.J., 1977. The oxygen isotope stratigraphic record of the Thompson, L.G., Mosley-Thompson, E., Davis, M.E., Lin, P.-N., Hen- Late Pleistocene. Philosophical Transactions of Royal Society of derson, K.A., Cole-Dai, J., Bolzan, J.F., Liu, K.-B., 1995. Late glacial London B280, 169}182. stage and Holocene tropical ice core records from HuascaraH n, Peru. Shackleton, N.J., 1987. Oxygen isotopes, ice volume and sea level. Science 269, 46}50. Quaternary Science Reviews 6, 183}190. Thouveny, N., Moreno, E., Delanghe, D., Candon, L., Lancelot, Y., Shackleton, N.J., Duplessy, J.-C., Arnold, M., Maurice, P., Hall, M.A., Shackleton, N.J., 2000. Rock magnetic detection of distal ice- Cartlidge, J., 1988. Radiocarbon age of last glacial Paci"c deep rafted debries: clue for the identi"cation of Heinrich layers water. Nature 335, 708}711. on the Portuguese margin. Earth and Planetary Science Letters 180, Shen, G.T., Dunbar, R.B., 1995. Environmental controls on uranium in 61}75. reef corals. Geochimica et Cosmochimica Acta 59, 2009}2024. Thunell, R.C., Anderson, D.M., Gellar, D., Miao, Q., 1994. Sea-surface Shen, C.-C., Lee, T., Wang, C.-H., Dai, C.-F., Li, L.-A., 1996. The temperature estimates for the tropical western Paci"c during the calibration of D(Sr/Ca) versus sea surface temperature relationship last glaciation and their implications for the Paci"c warm pool. fot Porites corals. Geochimica et Cosmochimica Acta 60, Quaternary Research 41, 255}264. 3849}3858. Tukey, J.W., 1970. Some further inputs. In: Merriam, D.F., (Ed.), Geos- Sikes, E.L., Keigwin, L.D., 1994. Equatorial Atlantic sea-surface tem- tatistics: a Colloquium. Plenum, New York. Y d perature for the last 30 kyr: a comparison of U, O and Waelbroeck, C., Labeyrie, L., Duplessy, J.-C., Guiot, J., Labracherie, foraminiferal assemblage temperature estimates. Paleoceanography M., Leclaire, H., Duprat, J., 1998. Improving past sea surface tem- 9, 31}46. perature estimates based on planktonic fossil faunas. Paleoceano- Sikes, E.L., Keigwin, L.D., 1996. A reexamination of northeast Atlantic graphy 12, 272}283. sea surface temperature and salinity over the last 16 kyr. Watkins, J.M., Mix, A.C., 1998. Testing the e!ects of tropical temper- Paleoceanography 11, 327}342. ature, productivity, and mixed-layer depth on foraminiferal transfer Sikes, E.L., Samson, C.R., Guilderson, T.P., Howard, W.R., 2000. Old functions. Paleoceanography 13, 96}105. radiocarbon ages in the southwest Paci"c Ocean during the last Weaver, A.J., Eby, M., Fanning, A.F., Wiebe, E.C., 1998. Simulated glacial period and deglaciation. Nature 405, 555}558. in#uence of carbon dioxide, orbital forcing, and ice sheets on the Sikes, E.L., Volkman, J.K., 1993. Calibration of long-chain alkenone climate of the Last Glacial Maximum. Nature 394, 847}853. unsaturation ratios for palaeotemperature estimation in cold polar Weaver, P.P.E., Chapman, M.R., Eglinton, G., Zhao, M., Rutledge, D., waters. Geochimica et Cosmochimica Acta 57, 1883}1889. Read, G., 1999. Combined coccolith, foraminiferal, and biomarker Sikes, E.L., Volkman, J.K., Robertson, L.G., Pichon, J.-J., 1997. Al- reconstruction of paleoceanographic conditions over the past kenones and alkenes in surface waters and sediments of the South- 120 kyr in the northern North Atlantic (593N, 233W). Paleoceano- ern Ocean: implications for paleotemperature estimation in polar graphy 14, 336}349. regions. Geochimica et Cosmochimica Acta 61, 1495}1505. Webb, R.S., Rind, D.H., Lehman, S.J., Healy, R.J., Sigman, D., 1997. Sinclair, D.J., Kinsley, L.P.J., McCulloch, M.T., 1998. High resolution In#uence of ocean heat transport on the climate of the Last Glacial analysis of trace elements in corals by laser ablation ICP-MS. Maximum. Nature 385, 695}699. Geochimica et Cosmochimica Acta 62, 1889}1901. Weber, J.N., 1973. Incorporation of strontium into reef coral Sonzogni, C., Bard, E., Rostek, F., Dollfus, D., Rosell-MeleH , A., Eglin- skeletal carbonate. Geochimica et Cosmochimica Acta 37, ton, G., 1997. Temperature and salinity e!ects on alkenones ratios 2173}2190. measured in surface sediments from the Indian Ocean. Quaternary Wefer, G., Berger, W.H., Bickert, T., Donner, B., Fischer, G., Kemle-von Research 47, 344}355. MuK cke, S., Meinecke, G., MuK ller, P.J., Mulitza, S., Niebler, H.-S., Sowers, T., Bender, M., 1995. Climate records covering the last deglaci- PaK tzold, J., Schmidt, H., Schneider, R.R., Segl, M., 1996. Late ation. Science 269, 210}214. Quaternary surface circulation of the South Atlantic: the stable Steig, E.J., Brook, E.J., White, J.W.C., Sucher, C.M., Bender, M.L., isotope record and implications for heat transport and productivity. Lehman, S.J., Morse, D.L., Waddington, E.D., Clow, G.D., 1998. In: Wefer, G., Berger, W.H., Siedler, G., Webb, D.J. (Eds.), The Synchronous Climate Changes in Antarctica and the North Atlan- South Atlantic: Present and Past Circulation. Springer-Verlag, tic. Science 282, 92}95. Berlin, pp. 461}502. A.C. Mix et al. / Quaternary Science Reviews 20 (2001) 627}657 657

Weinelt, M., Sarnthein, M., P#aumann, U., Schulz, H., Junk, S., Erlen- Yoshioka, S., Ohde, S., Kitano, Y., Kanamori, N., 1986. Behaviour of keuser, H., 1996. Ice-free Nordic seas during the Last Glacial Max- magnesium and strontium during the transformation of coral aragon- imum?. Potential sites of deepwater formation. Palaeoclimates 1, ite to calcite in aquatic environments. Marine Chemistry 18, 35}48. 283}309. Zahn, R., Mix, A.C., 1991. Benthic foraminiferal dO in the oceans Welling, L.A., Pisias, N.G., 1998. How do Radiolarian sediment assem- temperature-salinity-density "eld: Constraints on ice-age thermoha- blages re#ect the ecology of the surface ocean?. Paleoceanography line circulation. Paleoceanography 6, 1}20. 12, 131}149. Zahn, R., Pedersen, T.F., Bornhold, B.D., Mix, A.C., 1991. Water mass Wol!, T., Mulitza, S., RuK hlemann, C., Wefer, G., 1999. Response of conversion in the glacial subarctic Paci"c (543N, 1483W). Physical the tropical Atlantic thermocline to late Quaternary trade wind constraints and the benthic-planktonic stable isotope record. changes. Paleoceanography 14, 384}396. Paleoceanography 6, 543}560. Yin, J.H., Battisti, D.S., 2000. The importance of tropical sea surface Zaucker, F., Stocker, T.F., Broecker, W.S., 1994. Atmospheric fresh- temperature patterns in sumulations of Last Glacial Maximum water #uxes and their e!ect on the global thermohaline circulation. climate. Journal of Climate, in press. Journal of Geophysical Research 99, 12,443}12,457. Yokoyama, Y., Lambeck, K., De Deckker, P., Johnson, P., Zielinski, U., Gersonde, R., 1997. Diatom distributions in southern Fi"eld, K., 2000. Timing for the maximum of the Last ocean surface sediments (Atlantic sector): implications for paleoen- Glacial constrained by lowest sea-level observations. Nature 406, vironmental reconstructions: Palaeogeography. Palaeoclimatology 713}716. and Palaeoecology 129, 213}250.